WPD Dissertation

Investigation of Planar Power Dividers
for Wireless Communication Systems
by
Chen, Anqi
B.Eng.
A thesis submitted in accordance with the requirements for the award of
the degree of Doctor of Philosophy of the University of Liverpool
September 2018
Declaration
I hereby declare that except where specific reference is made to the work of others, the contents
of this dissertation are original and have not been submitted in whole or in part for consideration
for any other degree or qualification in this, or any other University. This dissertation is the
result of my own work and includes nothing which is the outcome of work done in collaboration,
except where specifically indicated in the text.
The copyright of this thesis rests with the author. Copies (by any means) either in full, or of
extracts, may not be made without prior written consent from the author. Copyright © 2018
Anqi Chen, all rights reserved.
Chen, Anqi
2018
P a g e | ii
Acknowledgements
First and foremost, I would like to express my deepest gratitude to my main supervisor Dr.
Zhou Jiafeng for granting me the precious opportunity to test myself at the highest level. You
have encouraged and helped me to grow up from a fresh graduate student to a professional
researcher. I learned so much from you both academically and personally. Thank you for the
invaluable comments and advice on my research as well as my boyfriend and career. It is a
great honour for me to be one of your PhD students, and I hope I have repaid the faith you
showed in me. You have taught me to be a rigorous researcher and a realistic engineer with
commercial awareness. I will always treasure your support, encouragement, and guidance.
I would also like to thank my parents, especially my father. You have always supported me
with no expectation of a reward. Your continuous help and understanding have made my life
full of love and I am grateful for everything you have done. I would also like to express my
appreciation to my boyfriend. I would never have succeeded without his tolerance, support,
encouragement, and patience. This work is dedicated to you.
Special thanks are also paid to my brilliant and lovely colleagues and friends; in particular to
Dr. Chaoyun Song, Dr. Qian Xu, Dr. Lei Xing, Dr. Sheng Yuan, Dr. Saqer S. Alja'afreh, Dr.
Muayad Kod, Dr. Chaoyun Song, Dr. Zhihao Tian, Dr. Zhouxiang Fei, Dr. Muaad Hussein, Dr.
Abed Pour Sohrab, Yuan Zhuang, Manoj Stanley, Wenzhang Zhang, Chen Xu, Tianyuan Jia,
Dr. Dajun Lei, Rui Pei, Zhenzhen Jiang, Jingchen Wang and Umniyyah Ulfa for many fruitful
discussions and enjoyable moments. Particular thanks should also be paid to Mark Norman,
John Owens, Gareth Blacoe and John Gillmore from the Electrical Workshop for always being
very kind to me and fabricating my circuits very quickly and beautifully.
P a g e | iii
Table of Contents
Table of Contents .................................................................................................................... iv

List of Figures .........................................................................................................................vii
List of Tables ........................................................................................................................... xi
List of Publications ................................................................................................................xii
Acronyms xiv
Abstract xvi
Chapter 1 Introduction ........................................................................................................ 1
1.1 Power Dividers............................................................................................................. 1
1.2 PD in Modern Wireless Communication Systems..................................................... 10
1.3 The Motivations of the Work ..................................................................................... 12
1.4 Thesis Organization ................................................................................................... 13
References ............................................................................................................................ 14
Chapter 2 A Review of Power Dividers............................................................................ 15
2.1 Broadband PD Technologies ..................................................................................... 15
2. 1. 1 PDs with Multi-section .......................................................................................... 15
2. 1. 2 PDs with Stepped-impedance Transformers .......................................................... 20
2. 1. 3 PDs with Coupled-Line.......................................................................................... 22
2.2 Reconfigurable PDs ................................................................................................... 29
2. 2. 1 Tunable Power Dividing Ratios ............................................................................. 29
2. 2. 2 Tunable Operating Frequency................................................................................ 32
2. 2. 3 Hybrid Reconfigurable PDs ................................................................................... 36
2.3 Coupled Resonator PDs ............................................................................................. 41
2.4 Millimetre-Wave PDs ................................................................................................ 46
2.5 Summary .................................................................................................................... 53
References ............................................................................................................................ 54
Chapter 3 Design of a Broadband Wilkinson Power Divider with Wide Range Tunable
Bandwidths 62
3.1 Introduction ................................................................................................................ 62
3.2 Design of the PD ........................................................................................................ 64
Return Loss ......................................................................................................... 65
Isolation Circuit .................................................................................................. 68
P a g e | iv
Design Parameters and Simulation ..................................................................... 69
3.3 Parametric Studies Regarding Bandwidth ................................................................. 72
Bandwidth Regarding 𝜽𝟏 ................................................................................... 72
Bandwidth Regarding Capacitance 𝑪𝒑 ............................................................... 76
3.4 General Cases............................................................................................................. 79
3.5 Experimental Results ................................................................................................. 80
Fabrication and Measurement ............................................................................. 80
Two-way PD ....................................................................................................... 82
Eight-way PD ...................................................................................................... 85
3.6 Summary .................................................................................................................... 88
References ............................................................................................................................ 89
Chapter 4 A Horst-type PD with Wide Frequency Tuning Range Using Varactors ... 91
4.1 Introduction ................................................................................................................ 91
4.2 Analytical Design Equation for the Proposed Horst-type PD ................................... 93
Even-mode Analysis ........................................................................................... 94
Odd-mode Analysis ............................................................................................ 96
4.3 Design Method ........................................................................................................... 98
Determination of Impedance 𝐙𝟏and 𝐙𝟐 ............................................................ 98
Frequency Tuning Range Regarding 𝑪𝟏 ........................................................... 99
Tuning range of the tunable PD ........................................................................ 102
Design Guidance ............................................................................................... 104
4.4 Fabrication and Measurement .................................................................................. 105
4.5 Summary .................................................................................................................. 110
References .......................................................................................................................... 111
Chapter 5 Millimetre-Wave CPW PDs Using Spiral Resonators for Miniaturization.
113
5.1 Introduction .............................................................................................................. 113
5.2 Analysis of the Coupled-Resonator PD ................................................................... 115
5.3 CPW PD with Spiral Resonators ............................................................................. 118
Resonant Frequency .......................................................................................... 119
Coupling Coefficient k ...................................................................................... 120
External Coupling Qe ........................................................................................ 122
5.4 Configurations and Designs ..................................................................................... 123
Proposed PDs .................................................................................................... 123
Splitting Frequencies (diplexer) ........................................................................ 126
5.5 Fabrication Process and Measurement..................................................................... 131
Page|v
5.6 Summary .................................................................................................................. 134
References .......................................................................................................................... 135
Chapter 6 Summary and Future Work ......................................................................... 138
6.1 Summary .................................................................................................................. 138
6.2 Key Contributions .................................................................................................... 139
6.3 Future Work ............................................................................................................. 140
References .......................................................................................................................... 144
P a g e | vi
List of Figures
Fig. 1. 1. A diagram of wireless communication networks [1]. ......................................................................... 1

Fig. 1. 2. Schematic of the resistive PD, (a) star format PD, (b) delta format PD. .......................................... 2
Fig. 1. 3. (a) E-plane waveguide, (b) H-plane waveguide and (c) Microstrip T-Junction. .............................. 4
Fig. 1. 4. Transmission line model of a lossless T-junction. ............................................................................... 5
Fig. 1. 5. N-way Wilkinson PD [6]. ...................................................................................................................... 6
Fig. 1. 6. (a) Signal split in a two-way Wilkinson PD, (b) schematic of a two-way Wilkinson PD. ................ 7
Fig. 1. 7. (a) Circuit schematic and (b) configuration of a conventional rat-race coupler [8]. ....................... 8
Fig. 1. 8. Basic Gysel topology. ............................................................................................................................ 9
Fig. 2. 1 (a) Schematic of three-section Wilkinson PD, (b) photo of industrial products. ............................ 16
Fig. 2. 2 (a) Even-mode circuit of the design in [34] based on the structure of two segments, (b) printed
circuit. ........................................................................................................................................................ 18
Fig. 2. 3 (a) Simulated and measured S11 and S21, (b) simulated and measured S22 and S23 [35]. ................ 19
Fig. 2. 4 Structure of stepped-impedance transformer. ................................................................................... 20
Fig. 2. 5 (a) Photograph of the designed four-way PD, (b) S11 S21 and output return losses and (c)
isolations of the designed four-way PD with ripple 15 dB [46]. ............................................................ 21
Fig. 2. 6 (a) Schematic model of coupled-line, (b) coupled-line side view. ..................................................... 22
Fig. 2. 7 Fabricated PDs of (a) conventional and (b) coupled-line design for a gap of 0.56 mm. ................. 24
Fig. 2. 8 Measured (a) S11, (b) S21, (c) S22, and (d) S23 comparing to the conventional Wilkinson PD [67]. . 25
Fig. 2. 9 (a) Schematic of proposed Ultra-wideband PD and (b) fabricated circuit. Simulation and
measured comparison of (c) S11, S21, and (d) S22 & S23 [68]. .................................................................. 26
Fig. 2. 10 (a) Schematic and (b) fabricated circuit of the proposed wideband filtering PD, simulation and
measured results of (c) S11, S22 and S33, (d) S21, S31and S23 [69]. ............................................................ 28
Fig. 2. 11 (a) Tri-coupled line structure, (b) layout and photography of the proposed tunable PD [70]..... 30
Fig. 2. 12 Measured (a) S21 and S31, (b) S11 and the phase difference between output ports, (c) S23 [70]. .... 31
Fig. 2. 13 (a) Configuration and (b) fabricated circuit of the PD. .................................................................. 32
Fig. 2. 14 Comparison of simulated and measured (a) S11, (b) S22, (c) S21, and (d) S32.................................. 33
Fig. 2. 15 (a) Cascaded a filter and a PD to form a filtering PD, and (b) multiple bandpass filter to form
tunable bandpass filter. ............................................................................................................................ 34
Fig. 2. 16 (a) Configuration and (b) fabricated circuit of the proposed PD in [72], and comparison of
measured and simulated performances regarding (c) S11, and (d) S22 & S33. ....................................... 35
Fig. 2. 17 Reconfigurable RF front end with tunable power dividing and controllable filtering for
antennas. .................................................................................................................................................... 36
Fig. 2. 18 (a) From cascaded structure to an integrated tunable filtering PD, (b) proposed design
configuration, and (c) fabricated circuit (0.24𝝀𝒈 ×0.12𝝀𝒈) for [73]..................................................... 37
P a g e | vii
Fig. 2. 19 Simulations and measurements results for demonstrating (a) different power division ratio, (b)
bandwidth tunability, (c) centre frequency tunability, and (d) cut-off edge tunability and in-band
differential phase of [73]. .......................................................................................................................... 38
Fig. 2. 20 (a) Circuit model of reconfigurable wideband filtering PD and (b) fabricated circuit for [74]. . 40
Fig. 2. 21 The (a) coupling scheme, and (b) the structure of the in-phase PD in [79]. .................................. 41
Fig. 2. 22 Selective band-edge regarding (a) L4 and (b) S4, and tuning frequency band regarding (c) L1,
and (d) W1, (e) comparison of simulated and measured results [79]. ................................................... 42
Fig. 2. 23 (a) Coupling scheme of the PD in [80], and the (b) quarter-wavelength unit-impedance
resonator, and the (c) configuration of the PD in [80]. .......................................................................... 44
Fig. 2. 24 Comparison of simulation and measurement in terms of (a) |S11| and |S21|, and (b) |S22|, |S33| and
|S23|. ............................................................................................................................................................. 45
Fig. 2. 25 (a) Modified Wilkinson PD and micro-photographs of the PD in [81], and measured results of
(b) transmission, and (c) return losses and isolation. ............................................................................. 47
Fig. 2. 26 Structure of SIW and its equivalent rectangular waveguide.......................................................... 48
Fig. 2. 27 (a) A prototype of the proposed HMSIW PD, and simulations and measurements for (b) S11, S21
and S31, and (c) S22, S33 [92]. ..................................................................................................................... 49
Fig. 2. 28 (a) Simplified schematic of the coplanar ring divider [91], (b) Simulated surface current flow at
80 GHz........................................................................................................................................................ 50
Fig. 2. 29 Measured results compared with Simulations of (a) S11, (b) S21, (c) S33 and (d) S23. .................... 51
Fig. 2. 30 (a) Structure of CPW ring PD and (b) S-parameters simulation and measurement comparison
of (b) S11, (c) S22, (d) SS21 and (e) S23 [95]. ................................................................................................ 52
Fig. 3. 1 The proposed broadband bandwidth tunable PD. ............................................................................ 64
Fig. 3. 2 Even-mode equivalent circuit for the proposed divider. ................................................................... 65
Fig. 3. 3 Odd-mode equivalent circuit for the proposed divider. .................................................................... 68
Fig. 3. 4 (a) AWR schematic circuit, (b) comparison of the simulated 20 dB bandwidth of S11 between the
proposed divider and previous works [1], [3] using AWR. ................................................................... 70
Fig. 3. 5 Fabrication layout of proposed two-way PD...................................................................................... 71
Fig. 3. 6 Circuit parameters for realising different additional zero-reflection frequencies. ........................ 73
Fig. 3. 7 The contour of S11 with frequency as a function of 𝜽𝟏 when 𝜽𝟐 = 𝛑𝟐 − 𝜽𝟏 CP = 0.256
pF 𝐚𝐧𝐝 𝜽𝟑 = 𝟕°. ........................................................................................................................................ 74
Fig. 3. 8 The contour of S11 with frequency as a function of 𝜽𝟏 when 𝜽𝟐 = 𝛑𝟐 − 𝜽𝟏, 𝜽𝟑 = 𝟕° and (a)
𝑪𝒑=0.2 pF, (b) 𝑪𝒑=0.4 pF and (c) 𝑪𝒑=0.8 pF. ........................................................................................ 75
Fig. 3. 9 Changing CP produces varying 𝐒𝟏𝟏 𝐚𝐧𝐝 𝐒𝟐𝟏 with 𝜽𝟏 = 𝟓𝟑°, 𝜽𝟐 𝟑𝟕° 𝐚𝐧𝐝 𝜽𝟑 = 𝟕°. .................... 77
Fig. 3. 10 The contours of S11 as a function of frequencies and CP when 𝜽𝟏 = 𝟓𝟑° and 𝜽𝟑 = 𝟕°. ............... 78
Fig. 3. 11 The maximum 20 dB bandwidth regarding 𝜽𝟏 and corresponding 𝑪𝒑. ....................................... 79
Fig. 3. 12 Photographs of the measurement set up and the fabricated two-way divider. ............................ 81
Fig. 3. 13 Measured results of (a) S11 & S21, (b) S22 & S23 for proposed two-way PD.................................... 83
Fig. 3. 14 Simulated results of (a) S11 & S21, (b) S22 & S23 for proposed two-way PD. .................................. 84
Fig. 3. 15 The fabricated circuit of eight-way PD. ........................................................................................... 85
P a g e | viii
Fig. 3. 16 (a) Measured response and (b) simulated results of S11, S21, S22 and S23 for the eight-way PD. .. 86
Fig. 3. 17 (a) S11, S41/S51, and (b) S44, S45 and S55 for ports 4&5, (c) S11, S61/S71, and (d) S66, S67 and S77 for
ports 6&7, (e) S11, S81/S91, and (f) S88, S89 and S99 for ports 8&9. .......................................................... 87
Fig. 4. 1 Structure of the proposed Horst-type PD. ........................................................................................ 93
Fig. 4. 2 The equivalent circuit of the proposed structure for even-mode analysis. ..................................... 94
Fig. 4. 3 The equivalent circuit of the proposed structure for the odd-mode analysis. ................................ 96
Fig. 4. 4 Theoretical reflection-zero frequency curve with changing Z1 and Z2. .......................................... 99
Fig. 4. 5 Theoretical S11 envelope curves as a function of 𝑪𝟏for different impedance of front
transmission lines for (a) α=1, (b) α=1.15, (c) α=0.85 and (d) α=0.6. .................................................. 101
Fig. 4. 6 The relationship between the worst in-band S11 and changing 𝐙𝟏................................................ 102
Fig. 4. 7 Theoretical required capacitance 𝑪𝟏 and 𝑪𝟐 versus corresponding frequency with 𝐙𝟐 =2 when
𝐙𝟏=0.4, 𝜽𝟏 = 𝟑𝟎° and 𝜽𝟐 = 𝟏𝟖°. ........................................................................................................... 104
Fig. 4. 8 (a) Configuration of the proposed PD and (b) photo of the fabricated PD. ................................ 105
Fig. 4. 9 Comparison between simulations and measurements, (a) S11 and (b) S21. ................................... 108
Fig. 4. 10 Comparison between simulations and measurements, (a) S22, and (b) S23. ................................ 109
Fig. 5. 1 Relative loss, size, and cost of various RF resonators [1]. ............................................................... 113
Fig. 5. 2 The equivalent circuit of a PD with three coupled resonators. ...................................................... 115
Fig. 5. 3 The configuration of the proposed CPW PD with three resonators. ............................................. 118
Fig. 5. 4 (a) The configuration of a quarter-wavelength spiral resonator, and (b) the relationship between
resonance frequencies and side-length L. .............................................................................................. 119
Fig. 5. 5 (a) Face-to-back oriented, and (b) back-to-back reversely oriented resonators. .......................... 120
Fig. 5. 6 Resonant mode splitting phenomena of (a) face-to-back oriented and (b) back-to-back reversely
oriented. (c) Coupling coefficient k versus the distance d between resonators. ................................. 121
Fig. 5. 7 (a) Feeding structure of the PD, and (b) relationship between external Q e and centre frequency f0
with the off-set Lfeed in (a). ...................................................................................................................... 122
Fig. 5. 8 (a) Feeding structure of the resonator, and (b) variation of Q e and centre frequency f0 with the
gap width in (a)........................................................................................................................................ 123
Fig. 5. 9 (a) Configuration of the optimized three-resonator PD with a 1:4 power division ratio, (b) 1:1
power division, (c) 1:2 power division and (d) 1:4 power division. ..................................................... 124
Fig. 5. 10 (a) Configuration of the CPW PD with five resonators, and (b) its simulated performance. .... 125
Fig. 5. 11 RF front end of a cellular base station [20]. ................................................................................... 126
Fig. 5. 12 The topology of the proposed resonator-based diplexer. .............................................................. 127
Fig. 5. 13 Theoretical response of (a) the dual-band bandpass filter, and (b) the bandpass filters of lower
and higher frequency bands, respectively [22]. .................................................................................... 128
Fig. 5. 14 (a) Proposed five-resonator diplexer, and (b) its performances. .................................................. 130
Fig. 5. 15 (a) Circuit layout of a 2 GHz PD, and (b) the fabricated circuit. ................................................. 131
Fig. 5. 16 Comparison between the simulated and measured results of the 2 GHz PD. ............................. 132
Fig. 5. 17 (a) Circuit layout of a 2 GHz diplexer, and (b) the fabricated circuit. ........................................ 132
P a g e | ix
Fig. 5. 18 Comparison between the simulated and measured (a) S11, S22 and S33, and (b) S21, S31 and S23 of
the 2 GHz Diplexer. ................................................................................................................................. 133
Fig. 6. 1 Resist pattern schematic by using EBL [2]. .................................................................................... 141
Fig. 6. 2 (a) The GSG probe, and (b) microphotograph of the fabricated ring divider under test [6], (c)
Typical 110 GHz System [7]. .................................................................................................................. 142
Fig. 6. 3 Configuration of an interdigital capacitor. ..................................................................................... 143
Page|x
List of Tables
Table 3. 1 COMPARISON OF CALCULATED AND OPTIMIZED PARAMETERS. ............................. 72

Table 3. 2 CALCULATED DESIGN PARAMETERS FOR DIFFERENT ADDITIONAL ZERO
REFLECTION FREQUENCIES............................................................................................................. 73
Table 3. 3 PERFORMANCE COMPARISON OF TWO-WAY WILKINSON-TYPE POWER
DIVIDERS WITH OTHER WORKS. .................................................................................................... 85
Table 4. 1 DESIGN PARAMETERS OF THE PROTOTYPE TUNABLE POWER DIVIDER. ............ 106
Table 4. 2 COMPARISON TO OTHER WORKS........................................................................................ 110
P a g e | xi
List of Publications
[1] A. Chen, Y. Zhuang, J. Zhou, Y. Huang and L. Xing, “Design of a Broadband

Wilkinson Power Divider with Wide Range Tunable Bandwidths by Adding a Pair of
Capacitors,” IEEE Transactions on Circuits and Systems II: Express Briefs ( Early
Access ), 2018.
[2] A. Chen, Y. Zhuang, Y. Huang and J. Zhou, ‘A Horst-type Power Divider with Wide
Frequency Tuning Range Using Varactors,’ IEEE Access, ISSN: 2169-3536， pp: 1-
13, 2018.
[3] A. Chen, B. Mousa, Y. Zhuang, Y. Huang and J. Zhou, “Compact Ka-band substrate-
integrated waveguide filter with spurlines for satellite communication systems,” 2016
IEEE 9th UK-Europe-China Workshop on Millimetre Waves and Terahertz
Technologies (UCMMT), Qingdao, China, 2017.
[4] A. Chen, Y. Zhuang, Y. Huang and J. Zhou, “Design of Reconfigurable Power Dividers
with Wide Tuning Ranges,” ARMMS RF and Microwave Society Conference, Oxford
(accepted), 2018.
[5] Y. Zhuang, Zhouxiang. Fei, A. Chen, Yi. Huang, K. Rabbi and Jiafeng. Zhou, “Design
of multiovtave high-efficiency power amplifier using stochastic reduced order models,”,
IEEE Transactions on Microwave Theory and Techniques., vol. 66, No.2, pp.1015-
1023, Feb, 2018.
[6] Y. Zhuang, C. Xu, C. Song, A. Chen, W. Lee, Y. Huang and J. Zhou, “A novel method
of improving current transformer-based energy extraction from AC power lines under
agnetic saturation,” IEEE Transactions on Industrial Electronics, 2018 (under review).
[7] Y. Zhuang, A. Chen, C. Xu, C. Song, Y. Huang and J. Zhou, “Position-free efficient
high power-transfer wireless power transfer using multiple bidirectional resonant coils,”
IEEE Transactions on Industrial Electronics, 2018 (submitting).
[8] Q. Xu, Y, Huang, L. Xing, Z. Tian, J. Zhou, A. Chen and Y. Zhuang, “Average
Absortion Coefficient Measurement of Arbitrarily Shaped Electrically Large Objects
P a g e | xii
in a Reverberation Chamer,” IEEE Transactions on Electromagnetic Compatibility, vol.
58, No.6, pp. 1776-1779, doi. 10.1109/TEMC.2016.2587679, 2016.
[9] Y. Zhuang, J. Zhou, Y. Huang and A. Chen, “Design of a broadband high efficiency
GaN power amplifier for GNSS applications,” IEEE MTT-S International Wireless
Symposium, Shanghai, China, 2016.
[10] Y. Zhuang, C. Xu, S. Yuan, C. He, J. Zhou, Y. Huang, A. Chen and W. W. Lee, “Design
of an Energy Harvesting System on Power Transmission Lines,” IEEE MTT-S Wireless
Power Transfer Conference, Taipei, Taiwan, May, 2017.
[11] Y. Zhuang, C. Xu, S. Yuan, C. He, A. Chen, W. Lee, Y. Huang and J. Zhou, “An
improved energy harvesting system on power transmission lines,” IEEE Wireless
Power Transfer Conference, June, 2017.
[12] Y. Zhuang, C. Xu, C. Song, A. Chen, W. Lee, Y. Huang and J. Zhou, “Using control
coil to improve magnetic energy extraction from AC power line under magnetic
saturation,” Asian Wireless Power Transfer Workshop. Dec, 2017.
[13] Y. Zhuang, C. Xu, A. Chen, W. Lee, Y. Huang and J. Zhou, “Distributed Matching
Network design for broadband power amplifiers,” IEEE Electrical Design of Advanced
Packaging and System Symposium, Dec, 2017. (Best Paper)
[14] Y. Zhuang, A. Chen, C. Xu, Y. Huang and J. Zhou, “Range-adaptive wireless power
transfer system using multi-antiparallel loops,” IEEE Asian Pacific Conference on
Antenna Propagation, Auckland, New Zealand, Aug, 2018. (Best Student Paper First
Prize).
P a g e | xiii
Acronyms
CAD Computer-aided Design
CDMA Code Division Multiple Access
CPS Symmetric Coplanar Strip-lines
CPW Coplanar Waveguide
CRLH Composite Right/Left Handed
DUT Device Under Test
EBL Electron Beam Lithography
GSG Ground-signal-ground
GSM Global System for Mobile Communications
HSPA High-Speed Packet Access
LRRM Line-Reflect-Reflect-Match
LTCC Lower Temperature Co-Fired Ceramic
LTE Long-Term Evolution
MHMIC Monolithic Hybrid Microwave Integrated Circuit
MMIC Monolithic Microwave Integrated Circuit
mm-Wave Millimetre-Wave
NA Network Analyzer
OFDM Orthogonal Frequency-Division Multiplexing
PCB Printed Circuit Boards
PD Power Divider
P a g e | xiv
RF Radio Frequency
RFID Radio Frequency Identification
Rx Receiver
SIW Substrate Integrated Waveguide
SOLR Short-Open-Load-Reciprocal
SOLT Short-Open-Load-Thru
TDMA Time-Division Multiple Access
Tx Transmitter
VNA Vector Network Analyzer
P a g e | xv
Abstract
This thesis presents several studies and designs on power dividers, filters and diplexers to
overcome respective challenges of these topics. This thesis focuses on the following three areas.
The first research area focuses on improving the bandwidth of Wilkinson-type power dividers.
For broadband power dividers design, the impedance matching is challenging. A novel design
method is proposed to broaden the bandwidth of Wilkinson-type power dividers by adding
capacitors in parallel with the main structures. Adding capacitors in parallel with transmission
lines is firstly proposed to control of the reflection zero frequency and operating bandwidth.
The fabricated circuit occupies a very compact circuit size at the same time. A two-way power
divider is designed and measured to validate the proposed method. The measurement indicates
that the device has a 15-dB bandwidth of 2.73:1. The proposed method can be applied to the
design of multi-way multi-section power dividers. An eight-way power divider is designed and
fabricated. The measured response has a bandwidth of 2.67:1. The proposed power divider
exhibits a much wider tunable bandwidth than other similar works. In addition, the design has
a simple layout, compact size, good physical and electrical isolation features.
The second area concentrates on reconfigurable power dividers. A novel power divider with a
wide operational frequency tuning range is presented and investigated. Using varactors in
parallel with transmission lines is firstly proposed to extend the operational frequency tuning
range. The reconfigurable feature is realized by electrically adjusting the bias voltage on
varactors. Theoretical formulas for the characteristic impedance and electrical length of the
transmission lines of the power divider are derived and analysed to precisely control the
operating frequency band. A power divider has been designed and fabricated to validate the
proposed design method. The measured results indicate that the power divider can achieve a
20-dB tunable operating frequency range of 4.63:1 which is the largest among all the other
similar works. Moreover, the power divider has a simple layout and a very compact size of
0.2𝝀𝒈 ×0.16𝝀𝒈 . This work demonstrates the excellent potential for working with frequency
agile components in modern communication systems, such as antenna arrays.
The last area under investigation is mm-Wave (Millimetre-Wave) coplanar power dividers
design. For mm-Wave applications, most of existing power dividers suffer from large size,
P a g e | xvi
high cost, complex fabrication processes and lossy performance. On these circumstances, CPW
power dividers using coupled resonators at V band are designed. This work has three main
contributions. The first is that the power dividing ratio of a power divider can be precisely
controlled via calculation. The second one is reducing circuit loss by replacing T- or Y-
junctions to eliminate air-bridges. The last contribution is circuit miniaturization by using spiral
resonators. Mathematical proof and parametric studies have been done in the thesis. A diplexer
is designed using this method which indicates an extensive applicability of the method for RF
components designs.
This thesis has successfully demonstrated novel design methods and techniques for power
divider designs. Several challenges including complex circuitry, large size and unnecessary
loss have been overcome by using the presented technologies. The research and knowledge in
this thesis should be of great significance to the future development of power dividers or even
other RF devices.
P a g e | xvii
Chapter 1: Introduction
Chapter 1 Introduction
Advanced researches are aiming to keep up with the social demands for a new generation of
wireless communication functions. As shown in Fig. 1. 1, the development of modern wireless
communication contributes to every aspect of life globally including health & safety, security
and wireless charging, etc. Radio Frequency (RF) devices, as the physical fundamentals for
signal transmission and receiving, play a vital role in wireless communication systems.
Fig. 1. 1. A diagram of wireless communication networks [1].
1.1 Power Dividers
Power dividers (PDs), also called power splitters, are one of the essential components in
building wireless communication systems. A PD can be used to split the power from the input
port into several smaller portions to different output ports. On the other hand, the PD can be
used for combining power from several input ports into one output port once the jobs of the
Page|1
input and output ports are reversed. The features and performance of the PD regarding
operating frequency, bandwidth, output ports isolations, circuit size etc. will significantly affect
the overall performance of the wireless communication system built on it. With the
development of communication systems, especially 5G communications, the spectral range for
data transmission is higher and higher. Multiple communication standards are desired due to
that a single standard transmission cannot satisfy the required performance of the system. In
that case, PDs are essential for assigning the signal to different paths. In RF systems and
measurement control systems, like power amplifier network, the performance of the power
distribution and coupling elements are crucial to the quality of the whole system. Many PDs
are designed to have different features to satisfy the requirements of a high-performance
communication system.
Resistive PD.
The easiest method to realise a PD is to use simple purely resistive elements [2]. Choosing
proper values of the resistors used in the PD can enable the input/output impedance of the
system to be constant when splitting or combining the power with the desired ratio. One of the
most common and fundamental forms of the PD is a three-way PD. The most adopted
configurations of the three-way resistive PD, namely star and delta format, are shown in Fig.
1. 2(a) (b) as follow.
(a) (b)
Fig. 1. 2. Schematic of the resistive PD, (a) star format PD, (b) delta format PD.
Page|2
For the star PD, the series resistors to the star centre are all equal. Since the input impedance
should be equal to the load impedance 𝑍0 and the resistance can be derived by
(𝑅 + 𝑍0 )⁄
𝑅+ 2 = 𝑍0
𝑍0⁄
𝑅= 3 (1.1)
Eq (1.1) shows that the resistance is one-third of the characteristic impedance of the ports. For
the delta format PD, the resistance of the resistors is identical to that of the characteristic
impedance of the RF system. With these simple forms of resistive PDs or splitters, any port
can be used as the input while the rest two severing as the outputs.
It should be mentioned that the resistive PD has double loss compared to a lossless PD’s
insertion loss. Taking the star format PD as an example, for an N-way star format PD, the
power dissipated in resistor A that is close to the common port is 𝑃𝑑𝑖𝑠𝑠𝑖𝑝𝑎𝑡𝑒𝑑_𝐴 = 𝑃𝑖𝑛 ×
(𝑁 − 1)⁄(𝑁 + 1), while the power dissipated in the two resistors B and C that are close to
output ports can be calculated as 𝑃𝑑𝑖𝑠𝑠𝑖𝑝𝑎𝑡𝑒𝑑_𝐵/𝐶 = 𝑃𝑖𝑛 × (𝑁 − 1)⁄[(𝑁 + 1) × 𝑁 2 ] . For
example, when N equals to 2, the output power is only 25% of the input power (-6.02 dB).
Furthermore, it is very convenient to achieve multi-ports PD by connecting more resistors to
the centre of the star. But the power loss will be higher, which is acceptable for some
applications. Resistive PDs are very easy to implement, and they can provide an extensive
operating bandwidth as the resistors are not frequency dependent. Also, a resistive PD is
relatively cheap. These factors make them very attractive for many applications if the main
disadvantage of power loss can be tolerated.
T-junction PD.
The T-junction [3]-[4] is a simple lossless three-port network that can be used for power
dividing or power combining. It can be formed in virtually any type of transmission line
medium (commonly used in the waveguide, microstrip or stripline structure). The lossless T-
junctions shown in Fig. 1. 3(a)-(c) can all be modelled as a junction of three transmission lines
as shown in Fig. 1. 4. 𝑍1 represents ports impedance, and 𝑍2 and 𝑍3 indicate load impedances
at Port 2 and Port 3.
Page|3
(a)
(b)
(c)
Fig. 1. 3. (a) E-plane waveguide, (b) H-plane waveguide and (c) Microstrip T-Junction.
Page|4
Fig. 1. 4. Transmission line model of a lossless T-junction.
Due to the fringing fields and higher order modes that associated with the discontinuity, stored
energy that can be regarded as a lumped susceptance, B. To have input impedance matched,
the following equation should be satisfied (1.2)
1 1 1
Yin  jB    (1.2)
Z2 Z3 Z0
𝐵 = 0 "not practical"
⇒ A lossless divider has mismatched ports
A transmission line is assumed to be lossless so that the characteristic impedances are all real.
In that case, B should be zero for perfect matching. However, B is not negligible in practice, so
a lossless divider has mismatched ports. But some reactive tuning element can usually be added
to the divider to cancel the susceptance B, at least over a narrow frequency range [5]. If not,
there will be poor isolation between the two output ports, and there will be a mismatch looking
into the output ports. The T-junction PD realised by the transmission lines will maintain a good
power division efficiency with a minimal loss.
Wilkinson-type PD.
Page|5
The Wilkinson N-way power splitter that shown in Fig. 1. 5 was invented by Ernest Wilkinson
[6] in 1960. The Wilkinson-type PD uses quarter-wavelength transformers, which are easy to
be fabricated using transmission lines on printed circuit boards (PCB). As a result, it offers the
possibility of a very cheap and straightforward splitter/divider/combiner while still providing
high levels of performance regarding loss and isolation between output ports.
Fig. 1. 5. N-way Wilkinson PD [6].
The N-way PD proposed by Wilkinson in [6] can split an input signal into N equal phase output
signals or combine N equal-phase signals into one in the opposite direction.
Looking at a two-way PD first, the main path of Wilkinson-type circuits relies on quarter-wave
transformers to match the split ports (output ports) to the common port (input port). Due to the
difficulty of achieving the perfect matching of a three-ports reciprocal network, usually a
resistor will be added between the output ports. The added resistor will not only contribute to
the terminations matching but also provides good isolation between the output ports. An equal-
amplitude, two-way splitting, single-stage Wilkinson is shown as Fig. 1. 6. The two split paths
are quarter-wave transformers of impedance √2Z0 . Therefore, these two paths provide the
same potential over the two ends of the resistor, the resistor consumes no resistive power, so
an ideal Wilkinson splitter is 100% efficient. Although Wilkinson PDs can be used as N-way
PDs, it is difficult to realise a planar structure since all ports need to connect to a common point
with a resistor.
Page|6
(a)
(b)
Fig. 1. 6. (a) Signal split in a two-way Wilkinson PD, (b) schematic of a two-way
Wilkinson PD.
A two-way Wilkinson PD is relatively simple and can be realised by different technologies

such as printed components on a printed circuit board (PCB) and lumped inductor and capacitor
elements [7]. From the power loss point of view, if perfect elements are adopted, the Wilkinson
PDs would not present any extra power loss from the division of the power. In practice, the
real components used for the Wilkinson splitter can be of low loss, especially when PCB
transmission lines are used along with a low loss PCB substrate material.
Furthermore, contributed by the isolation circuit realised by resistors, high-level isolation is

introduced between the output ports which will prohibit the ports from interfering with each
other. On the other hand, the realisation of the Wilkinson PD is highly dependent on the quarter
wavelength transmission lines. Once the operating frequency deviates from the original one, it
will cause a vital impedance mismatch, and limit the performance. Thus, the traditional
Wilkinson PDs are only suitable for relatively narrow-band applications.
Rat-race Coupler.
Page|7
The Rat-race coupler is one of the 180° hybrid couplers that are widely used in PD and
combiner designs. The schematic of a conventional rat-race coupler is outlined in Fig. 1. 7[8].
It is a lossless reciprocal four-port network. It comprises three 90° branches and one 270°
branch. In many references, the 270° branch is always decomposed into three 90° sections to
simplify the design. The characteristic impedance of the ring should be √2 times the
characteristic impedance of the port terminations with the purpose of impedance
(a)
(b)
Fig. 1. 7. (a) Circuit schematic and (b) configuration of a conventional rat-race coupler
[8].
Page|8
matching for all ports. As a PD, the rat-race coupler can be used for in-phase operation and
180° out-of-phase operation. For the case of the in-phase operation, a signal applied to port 1
is divided equally into Ports 2 and 3 with the same phase shift, and Port 4 is isolated. Conversely,
for out-of-phase operation, a signal injected into Port 4 is divided evenly into Ports 2 and 3
with 180° phase difference, and Port 1 is isolated.
The rat-race coupler is easily realisable and can be implemented in numerous technologies,
such as microstrip line, stripline and waveguide (magic-tee). Advantages are the simple design
and high degree of isolation between the input ports. However, it also has serious drawbacks,
such as relatively narrow bandwidth and sizeable occupied area due to the requisite 270°
transmission line section [8].
Fig. 1. 8. Basic Gysel topology.
Gysel PD.
Ulrich Gysel [9] proposed a new structure (shown in Fig. 1. 8) of PD in 1975. The configuration
is similar to a combination of a Wilkinson and a rat-race coupler [10]. The main advantage of
the Gysel power splitter is its power handling capacity. In a Wilkinson splitter, the resistor is
embedded into the network and must provide a short phase length for the scheme to work. The
Page|9
terminations in a Gysel are equal to Z0 and can be high-power loads if power handling is a
requirement. The Gysel divider is often used in kilowatt-level power combining, for example,
if some redundancy in a 50,000-Watt television transmitter is demanded a five-way Gysel
combiner with 15,000 Watt tubes could be use and it is able to remove one of the tubes for
service or replacement without taking the transmitter off-line.
1.2 PD in Modern Wireless Communication Systems
The increasing number of frequency bands and spectrum fragmentation demand the system
having broadband features. Modern wireless communication systems are evolving and
applying a multiplicity of emerging communication standards such as high-speed packet access
(HSPA), long-term evolution (LTE) and microwave access (WiMAX). These techniques
require different centre frequencies, bandwidths and modulation schemes [11]. Wireless
service providers have to offer a large number of radios for each standard resulting in a very
costly overall network infrastructure. To decrease costs, the flexibility and reusability of RF
transmitters and transceivers should be improved to allow the combination of different
technologies in a single piece of hardware. Driven by the demands of the consumer for wireless
services such as portable devices, smartphones and Internet access with the high data rate, the
industry and academia have responded and offered improved services and standards [12] [13].
For example, the first generation named 1G cellular was introduced in the early 1980s, based
on the analogue network operated at 900MHz. Then, the 2G cellular system based on Time-
Division Multiple Access (TDMA), Global System for Mobile Communications (GSM), and
Code Division Multiple Access (CDMA) was announced ten years later, which provided much-
improved data rate. Till now, the 4G using OFDM technology is operated around the world
offering high quality and enhanced functionality services. Nowadays, the idea of the 5G system
has been proposed to deliver higher data rate services. It will demand broader bandwidths and
higher efficiency with lower cost and smaller size devices. In this case, the design of broadband,
multi-band tunable, high isolation and compact PD attracts attention and becomes the research
topic.
In general, the main preferred functionalities of a PD can be summarized as the following

aspects.
P a g e | 10
 Broadband technology has drawn much attention to wideband antennas and antenna
arrays [1]. As an essential part of these applications, PDs need to keep up with the
spectrum extension.
 Many wireless communication systems have multi-band operating ranges. GSM works
at 0.9 GHz and 1.8 GHz; WLAN system works at 2.4/5.2/5/8 GHz. Many RF devices
tend to have a multi-band or reconfigurable feature to make the most of existing
spectrum and infrastructure sources.
 Nowadays, people demand mini portable products, like smaller and thinner mobile cell
phones and RFID (Radio Frequency Identification). Circuit miniaturisation and
integration are the trends. Obtaining compact sizes and maintaining high performance
of the device will be the primary challenge.
From the functionality point of view, there are PDs with different frequency configuration
which includes dual-band, multi-band and wideband PDs. Those types work for multi-standard
communication systems. In addition to multi-band or broadband features, PDs having tunable
operating frequencies are valuable for constructing multi-standard systems. Despite tuning
operating frequencies, power ratio can also be tuned. For some applications like antenna arrays,
and power amplifiers, power division ratios need to be customized. In those cases, PDs with
different power division configurations are desirable. Although most of the time the word
“power divider” indicate in-phase power division, arbitrary phase differences are required in
many applications like phased arrays.
In addition to the pre-mentioned types, function–integrated (hybrid) PDs are also very popular,
such as PDs with different termination impedances, PDs with filtering response and PDs with
harmonic suppression. There are several hybrid PD examples demonstrated in the next chapter.
Usually, those multi-functional PDs are suitable for systems that are very complex and highly
integrated to save space and avoid unnecessary loss.
From the fabrication technology’s aspect, there are conventional transmission lines PDs
including microstrip, stripline, coplanar waveguide, etc. It is usually easy to calculate the
dimensions and fabricate. However, to achieve advanced performance, unconventional
transmission lines are widely used, like coupled-line PDs and PDs using right-/left-handed
(CRLH) transmission lines and Lower Temperature Co-fired Ceramic (LTCC). The structures
P a g e | 11
of above PDs follow the fundamental topology like Wilkinson type PDs. The substrate
integrated waveguide (SIW) is also a popular technology. By using metallic via holes to prevent
signal leakage a “planar waveguide” can be formed. An SIW has waveguide properties but
without a bulky size or heavyweight.
1.3 The Motivations of the Work
In recent years, with the rapid development of wireless communication systems, many ideas
and technologies have been applied to improve the overall performance of the systems. Passive
PDs, as an essential component of RF modules, have attracted more and more interests in
researching and industry field. The nature of a three-port network determines that it cannot be
simultaneously lossless, reciprocal, and matched at all ports. As a result, pursuing optimised
performance for a PD is worth investigating. Nowadays, PDs have already gone far beyond
original prototypes to keep up with the innovation of modern communication technologies. For
practical use, people care about the performances, functions and fabrication processes of a PD.
State-of-the-art PDs exhibit improvement in various aspects according to different applications
and specifications. Subsequently, function-integrated PDs attract more and more attention
because this is the trend to have compact integrated systems.
To summarise, more works can be done to improve the performance of a PD. The main
objectives of this research are listed below:
 To study the wideband PD technologies and investigate the mechanisms of these

methods.
 To design a wideband PD with a simple layout and compact size. Analytical solutions
for determining the design parameters should be produced.
 To further study the mechanism of designing reconfigurable PDs and evaluate the pros
and cons of each method. To investigate multi-functional PDs.
 To develop a reconfigurable PD with a wide centre frequency tuning range and has a
very compact size.
 To explore mm-Wave PD designs and fabrication technologies.
P a g e | 12
 To design an mm-Wave coplanar waveguide PD using coupled resonators for easy

fabrication and reduction of conductor loss at such high frequencies.
1.4 Thesis Organization
The content of this thesis are organized in the following manner.
Chapter 1 provides the background of the project, the motivations of the work and the outline
of the thesis.
Chapter 2 is the literature review of previous works on PDs. The main focuses of the literature
review are broadband PDs, reconfigurable PDs and mm-Wave PDs.
Chapter 3 presents a wideband PD design. The procedures and methodology of the design are
provided. Measurements were taken to validate the design.
Chapter 4 introduces a reconfigurable PD with a wide centre frequency tuning range. The
tuning mechanism is explained, and the relationship between the varactors and the tunable
centre frequency is analysed in detail. Measurements were taken for validating of the design.
Chapter 5 explores mm-Wave CPW PDs using coupled resonators. The size can be
significantly minimized by using G-type spiral resonators. Also by the use of a common
resonator to replace a T-junction or Y-junction, the size can be further reduced. Moreover, by
controlling the coupling between the resonators, a different power dividing ratio can be
achieved. Based on the coupling method, a diplexer can also be implemented by controlling
the size and distances of resonators.
Chapter 6 concludes the works. The main objectives and contributions of the work are reviewed
and highlighted. Moreover, the potential extensions of the work and its challenges are presented.
This chapter has also listed several future research topics worthy of investigation.
P a g e | 13
References
[1] Hilink.com.cn. (2018). Reactive Power Splitter-Power Splitter-Manufacturer of Passive Products.

[online] Available at: http://www.hilink.com.cn/en/Power_divider/show/28.html [Accessed 17
Sep. 2018].
[2] Adams, Greg. "Designing Resistive Unequal Power Dividers." High-Frequency
Electronics (2007): 48-50.
[3] C. J. R. Smith and H. H. Sigmarsson, “Microstrip T-junction power divider with exponentially
tapered transmission lines,” IEEE Microw. Wireless Compon. Lett., vol. 26, no. 12, pp. 987–989,
Dec. 2016.
[4] G. R. Branner, B. P. Kumar, and D. G. Thomas, Jr., “Design of microstrip T junction power divider
circuits for enhanced performance,” in IEEE Proc. 38th Midwest Circuits Syst. Symp., 1995, vol.
2, pp. 1213–1215.
[5] D. M. Pozar, “Power dividers and directional couplers,” in Microwave Engineering, 4th ed.
Hoboken, NJ, USA: Wiley, 2012, ch. 7, pp. 362–364.
[6] E. Wilkinson, “An N-way hybrid power divider,” IEEE Trans. Microw. Theory Thech., pp. 116-
118, 1960.
[7] Djordjevic, A. Scattering Parameters of Microwave Networks with Multiconductor Transmision
Lines: Software and User's Manual. Artech House, 1990.
[8] H.-X. Xu, G.-M. Wang, and K. Lu, “Microstrip rat-race couplers,” IEEE Microw. Mag., vol. 12,
no. 4, pp. 117–129, Apr. 2011.
[9] U. H. Gysel, “A new N-way power divider/combiner suitable for highpower applications,” in IEEE
MTT-S Int. Microw. Symp. Dig., May 1975, pp. 116–118.
[10] J. Reed and G. Wheeler, “A method of analysis of symmetrical four-port networks,” IRE Trans.
Microwave Theory Tech., vol. MTT-4, pp. 246-252, Oct. 1956.
[11] Subharthi Paul, Long Term Evolution (LTE) & Ultra-Mobile Broadband (UMB) Technologies for
Broadband Wireless Access, [online] available at: http://www.cse. wustl.edu/~jain/cse574-
08/ftp/lte.pdf [Accessed 15 Oct. 2018]
[12] Jaiswal, S., Kumar, A. and Kumari, N., 2014. Development of wireless communication networks:
From 1G to 5G. International Journal of Engineering and Computer Science, pp 6053-6056.
[13] Sharma P. “Evolution of mobile wireless communication networks-1G to 5G as well as future
prospective of next generation communication network.” International Journal of Computer
Science and Mobile Computing. 2013 Aug;2(8):47-53.
P a g e | 14
Chapter 2: A Review of Power Dividers
Chapter 2 A Review of Power Dividers
In this chapter, a number of state-of-the-art planar PDs are introduced and discussed. The
innovations of these PDs lie in their performance breakthrough, physical configurations, and
function integrations. This chapter provides a basis for the development of planar PDs.
2.1 Broadband PD Technologies
PDs with different frequency configurations are quite important for modern communication
systems that require multi-standards. Since conventional PDs [1] [2] can not provide sufficient
bandwidth, many works have been reported on dual-band [3]-[18], multi-band [19]-[20], and
broadband PDs [21]-[28]. Compared to multi-band designs, PDs with wide frequency bands
are more desired because they are not only able to meet the requirements of the multi-standard
system, but can also contribute high-speed data rate [29]. It is reported in the literature [30] that
there are several techniques that identify the broadband PDs. The pros and cons of those
techniques will be analysed and discussed in the following sections with some typical design
examples. The discussion of the bandwidth in this thesis will be based on the scattering
parameters Sij.
2. 1. 1 PDs with Multi-section
Theoretically, the conventional single-section two-way Wilkinson-type PD has a usable

bandwidth (20-dB return loss) of 1.44:1. The impedances are perfectly matched at all ports that
generate reflection zeros at the centre frequency and its harmonic frequencies. However, Sii
and Sij between each adjacent reflection zeros are insufficient to make it a wideband. As a result,
the operational band only consists of one reflection zero. To broaden the operational bandwidth,
the reflection zeros can be allocated closer to make Sii and Sij sufficient (ideally all below 20
dB) for a wideband response. This is done by carefully choosing the impedances and lengths
of the transmission lines. In that case, the operational band will consist of several reflection
zeros. This cannot be accomplished by using conventional single-section Wilkinson PDs. On
P a g e | 15
the other hand, adapting multi-section paths to the main circuit, as shown in Fig. 2. 1(a), is a
comprehensive method to realise the broadband response.
(a)
(b)
Fig. 2. 1 (a) Schematic of three-section Wilkinson PD, (b) photo of industrial products.
The multi-section method was introduced in 1968 [31] to further broaden the operational
bandwidth of the Wilkinson PDs. Each section consists of a pair of transmission lines and a
resistor. This is done so that the circuit is like several single-section two-way Wilkinson PD
cascaded end to end. The multi-section structure can provide a significant increase in
bandwidth. More specifically, bandwidth can be increased to 2:1 with two sections, 4:1 with
four sections, and even 10:1 with seven sections. Theoretically, the bandwidth of multi-section
PDs can be extended without limitation with infinite sections. However, the size will be
sacrificed accordingly as shown in Fig. 2. 1(b). As shown in [30], the circuit is composed of a
P a g e | 16
finite number of resistors and the same transmission lines, and so synthesis for optimum
performance can be concluded. The design procedure is easy for two sections PDs, but
calculation difficulties of PDs with more sections will increase exponentially. Although in the
following half-century, many works [32]-[34] have been conducted on optimising the
algorithms with the help of simulation software, the designs are reliant on the approximations
and software tuning processes. In addition to the calculation difficulties, the size of a multi-
section Wilkinson PD is multiplied because the length of each section is usually a quarter
wavelength.
In summary, the advantages are that this multi-section solution can effectively provide
unlimited bandwidth, and the bandwidth in terms of design parameters can be calculated and
predicted. However, the longer transmission lines for the wider bandwidth increased the circuit
size which in turn brings more insertion loss.
To save space of the circuit, Yu [35] has analysed broadband Wilkinson PDs based on a
segmented structure formed by many transmission line segments in shunt with grounded
capacitors and series resistor-capacitor networks. The word “segment” is to distinguish “multi-
section”; while the latter indicates a multiple quarter-wavelength transmission line section,
“segmented” implies the total length of transmission lines is 𝜆⁄4 which downsizes the circuits.
The layout of a segmented structure is shown in Fig. 2. 2. The reason that this work is put in
“multi-section” is because the design still needs a number of resistors to connect each segment.
The operation bandwidth of Wilkinson PDs with a fixed transmission line length can be
extended by replacing the original quarter-wave transmission line section with the proposed
impedance transformer segmented. As shown in Fig. 2. 3, the measured bandwidth is 1 in this
work, which is a very significant improvement compared to the previous best-reported
bandwidth of 2.03:1 in [18].
P a g e | 17
(a)
(b)
Fig. 2. 2 (a) Even-mode circuit of the design in [34] based on the structure of two segments,
(b) printed circuit.
P a g e | 18
Fig. 2. 3 (a) Simulated and measured S11 and S21, (b) simulated and measured S22 and S23
[35].
Theoretically, more implemented segments can achieve larger bandwidths. Practically, though,
the limitations of the maximum achievable bandwidth depend on the implementation of the
required components values. In addition, too many lumped elements will consume more energy.
P a g e | 19
2. 1. 2 PDs with Stepped-impedance Transformers
Fig. 2. 4 Structure of stepped-impedance transformer.
Comparable to the multi-section solution, the use of stepped-impedance lines is also an easy
and accessible method to achieve broadband PDs. Stepped-impedance lines, also known as
stepped-impedance transformers, were initially proposed by Young [36] in 1962. The stepped-
impedance transformer is shown in Fig. 2. 4. It consists of several lossless transmission line
sections with electrical length φ and terminated in resistive load Z0 and ZL. The input
impedance looking into the transformer can be calculated by continually integrating the input
impedance of lossless transmission line formula (2.1) [36]:
Z L  jZ 0 tan(  l )
Zin (l )  Z 0 (2.1)
Z 0  jZ L tan(  l )
2𝜋
where phase constant β = and 𝜆 indicates the wavelength. Stepped-impedance transformers
𝜆
are widely applied in filter designs for circuit size reduction [37]-[40]. For Wilkinson PDs,
output ports need to be close in order to minimize phase shift across the resistor. However, an
extra 50 Ω transmission lines are needed to connect the resistor to the output ports. The stepped-
impedance transformers utilize the space occupied by 50 Ω interconnection lines to improve
impedance matching on all ports, in turn reducing the circuit size simultaneously. Refs [41]-
[43] provide some mathematical solutions of choosing the values for these transmission lines.
P a g e | 20
(a)
(b) (c)
Fig. 2. 5 (a) Photograph of the designed four-way PD, (b) S11 S21 and output return losses
and (c) isolations of the designed four-way PD with ripple 15 dB [46].
Papers [44]-[45] propose an impedance-matching method of interconnecting transmission lines

between adjacent stages to achieve wideband feature. Ref [46] has demonstrated a
comprehensive design of stepped-impedance transformers by choosing different parameters at
different stages to have concurrent impedance matching between stages, broadening the
bandwidth significantly. As shown in Fig. 2. 5(a), by cascading three single-section two-way
PDs into a four-way one, the measured results demonstrate a bandwidth of 3:1 as Fig. 2. 5(b)
(c) indicate. This method employs Chebyshev polynomials [47]-[49] to the interconnecting
P a g e | 21
transmission lines between stages. Theoretically, for the given number of stepped-impedance
transformers, the optimised design parameters can indeed be calculated. This method could be
very effective when dealing with conventional multi-stage, multi-way Wilkinson PDs with
plenty of interconnecting transmission lines, but these interconnecting transmission lines will
inevitably occupy more space.
2. 1. 3 PDs with Coupled-Line
(a)
(b)
Fig. 2. 6 (a) Schematic model of coupled-line, (b) coupled-line side view.
Coupled-line is one type of transmission line that is widely used in microwave integrated circuit
designs. An example of two microstrip coupled-lines over a common plane as shown in Fig. 2.
6(a) (b). Coupling will gradually emerge when they are sufficiently close to each other. The
coupling strength between the coupled lines can be modelled by introducing a mutual
inductance and capacitance per unit length, 𝐿𝑚 and 𝐶𝑚 . When 𝐿𝑚 = 𝐶𝑚 = 0, they are reduced
to the uncoupled equations describing the isolated individual lines.
Coupled-line resonators are widely used in wideband filter designs (see works in [50]-[55]).
As a result, these types of PDs are typically designed by following the filter design synthesis
P a g e | 22
which can generate finite transmission poles in the band, just as filters are performed. The
transmission poles and bandwidth can therefore be predicted and controlled according to the
filter performance. Moreover, coupled-lines and microstrip-slotline transitions [56]-[59] can
be regarded as transmission line transformers, which is a benefit for size reduction and is
capable for further bandwidth extension.
Conventional two-way Wilkinson PDs consists of two quarter-wavelength lines and an

isolation resistor. For high-frequency applications, chip resistors with small dimensions are
widely used for the isolation resistors. In such cases, the coupling between the two quarter
wavelength transmission lines can not be neglected. To avoid unwanted coupling, some
measures are taken to lead the designs for PDs [60]-[64]. However, as the analysis in [65]
illustrates, the broadband feature can be achieved with ports impedance matching at a relatively
wide frequency range by controlling the coupling between coupled-line. The coupled-line PD
can provide a filtering response compare to PDs with conventional transmission lines. In
addition, owing to the coupled-line structure, the width and spacing of the microstrip lines are
reduced, thereby resulting in the miniaturisation of circuit width. Since the first coupled-line
Wilkinson PD has been proposed in [66], there has been a worldwide interest to develop more
versatile components by using this method.
The PD in [67] utilises the coupling between transmission lines to enhance the bandwidth,
reducing the size of a convention Wilkinson PD. The two transmission lines of the conventional
Wilkinson PD are placed apart to avoid coupling between them, as unwanted coupling will
lead to diminished performance. However, for the design in [67], the two transmission lines
are close to each other because the coupling is a part of the impedance matching network.
P a g e | 23
(a) (b)
Fig. 2. 7 Fabricated PDs of (a) conventional and (b) coupled-line design for a gap of 0.56
mm.
Fig. 2. 7 exhibits the size comparison between a conventional Wilkinson PD and coupled-line
PD, and the size of the latter is a quarter of the former. The measurement results in Fig. 2. 8
indicate that the proposed design achieves slightly wider S11 bandwidth compared to the
conventional Wilkinson one. The circuit is analysed by using even- and odd-mode analyses.
The odd-mode impedance of the transmission line 𝑍0 will only change the matching output
ports while not affecting the matching condition for the input port. The 20-dB S11 bandwidth
is 1.45:1, and S22 and S33 are 2.08:1 which shows little improvement to the conventional
Wilkinson one. The bandwidth improvement is limited as only one coupled-line section is used
as shown in Fig. 2. 8(a). It should be noted that the isolation S23 and output return loss S22/S33
have visible shifts from the centre frequency. This is because the odd-mode impedance 𝑍0 will
change along with the frequency, which degrades the matching of the output ports. The
frequency shift can be mitigated by adding more lumped elements and transmission lines
between the output ports. This design shows the potential of the coupled-line method to change
the impedance matching at ports.
P a g e | 24
(a) (b)
(c) (d)
Fig. 2. 8 Measured (a) S11, (b) S21, (c) S22, and (d) S23 comparing to the conventional
Wilkinson PD [67].
2.1.3 Hybrid and Multi-Functional Wideband PDs
As the basic design in [67] can only produce small improvements in terms of bandwidth, more
technologies are expected to be combined to provide better performance such as wider
bandwidth and higher out-of-band selectivity. A better-designed case is presented in [68] as
shown in Fig. 2. 9(a), connecting coupled-line and stepped-impedance stubs to the
conventional Wilkinson PD. From Fig. 2. 9(a) (b), this design utilises a pair of quarter-
wavelength coupled lines that are located between one side of the resistor and the outputs. In
P a g e | 25
(a) (b)
(c) (d)
Fig. 2. 9 (a) Schematic of proposed Ultra-wideband PD and (b) fabricated circuit.

Simulation and measured comparison of (c) S11, S21, and (d) S22 & S23 [68].
addition, a pair of stepped-impedance open-circuited stubs and parallel coupled lines are
introduced to the output ports. As a result, four additional transmission poles can be generated.
Thus, impedance matching can be achieved in the overall ultra-wide band range if these poles
are properly allocated. The work can achieve a 3.4:1 13-dB bandwidth corresponding to a
bandwidth from 3.1 GHz to10.6 GHz which is relatively wide for PDs designed in X-band.
The advantages of the design are the simple structure (only one segment of Wilkinson divider),
the sharp and controllable reflection-zeros on S21 band edges, and the relatively wide frame.
P a g e | 26
Moreover, the isolation can be sufficient without adding more lumped elements to the isolation
circuit. However, this design requires a large area (four quarter-wavelength × two quarter-
wavelength) which will consume more energy, especially in the higher frequency bands.
A multi-functional, ultra-wideband PD using coupled-line is reported in [69]. This design

proposed a two-way dual-section PD with three pairs of coupled lines as shown in Fig. 2. 10.
This filtering PD can realise multiple properties, including good in-band return loss, high
isolation, good harmonic suppression, ultra-wide isolation frequency band, and ultra-wide
stopband. In Fig. 2. 10(a) (b), the quasi-coupled lines are composed of one transmission line
and three coupled lines for power splitting. By shifting the isolation circuit Ziso to the right-
hand side of the middle-coupled line, a broad isolated bandwidth between two output ports can
be observed. Also, with a weak–strong–weak coupling scheme and low impedance of the
middle-coupled line, the adopted quasi-coupled lines would result in a wide stopband.
Furthermore, unlike the shifted isolation band in [67], the work in [69] has already improved
the output ports matching by connecting the resistor R and the capacitor C in series. A reflection
zero would appear in the passband so that there would be conspicuous isolation. The open and
short stubs are used for harmonic suppression. Consequently, with the assistance of open and
shorted stubs, available passband frequencies at the input port can be significantly increased
and the isolated frequency band between two output ports can be broadened.
A design guideline is provided; by determining the centre operating frequency, the length of
coupled lines can be confirmed and then the even- and odd-mode impedances of coupled lines
can be calculated accordingly. Finally, a formula concludes the total impedance of the isolation
circuit; by substituting all design parameters of coupled lines, the impedance Ziso can be
calculated as a complex number while the real part is the resistance and the imaginary part can
be capacitance in the isolation circuit. Overall, as shown in Fig. 2. 10(c) (d) the design exhibits
a 15-dB bandwidth of 3.2:1 and a stop bandwidth of 3.04:1. This design not only successfully
demonstrates a wideband response for operating band, but also realizes an ultra-wide stopband
(4.5GHz to 14.5 GHz). It is worth noting that the fabricated circuit is very
P a g e | 27
(a)
(b)
(c) (d)
Fig. 2. 10 (a) Schematic and (b) fabricated circuit of the proposed wideband filtering PD,
simulation and measured results of (c) S11, S22 and S33, (d) S21, S31and S23 [69].
P a g e | 28
compact (0.2λ𝑔 x 0.15λ𝑔 ) compared to other works with similar performance. Normally,
compact size is referring to the circuit that the longest dimension is shorter than N quarter
wavelengths (𝑁 ×0.25λ𝑔 , N is the order of the circuit) at the centre operating frequency. This
type of multi-functional work is desired in many applications to replace filters.
2.2 Reconfigurable PDs
If a system requires changing operating frequencies but not very wideband performance, then
reconfigurable PDs are desired. Reconfigurable PDs are capable of selecting the frequencies
by adapting a tuning process. The multi-standard or self-adapted systems require tunable PDs
or reconfigurable PDs. The tunability of a PD mainly focuses on two aspects; one is the power
division ratio, and the other is operating frequencies.
2. 2. 1 Tunable Power Dividing Ratios
Plenty of applications such as antenna feeding networks and power amplifiers prefer small PDs
with highpower division ratios. These applications propose several challenges for PDs with
different dividing ratios. One is that the circuit is expected to be compact, and the arbitrary
power division ratio is best to fit in multi-standard communication systems. Typically, there
are three common methods to achieve unequal power division ratios, namely the uneven-
impedance method, the phase-modification method, and the hybrid method. The uneven-
impedance method is based on the notion that more power is passed through the path with
smaller impedance. Tuning the charateristic impedances of each ouput port will then achieve
the desired power division ratio easily. However, changing the characteristic impedances will
lead to extremely narrow or wide transmission lines, which are difficult to fabricate. The
second method, tuning the phase of the transmission lines instead of impedance, is easier.
Moreover, this method is capable of constructing high power division ratios. The third method,
a hybrid, is where the phase and impedance of the transmission line is modified simultaneously.
To realize single-band PDs with various functions, such as equal/unequal power splitting and
tunable power splitting ratio, conventional three-line coupled structure has been widely utilised.
The work in [68] shows a two-way PD with arbitrary power division as shown in Fig. 2. 11(a)
(b). The hybrid method is used in this design. A quarter-wavelength three-line coupled structure
P a g e | 29
is employed. One terminal of the central coupled-line is connected to the input port, whereas
the other terminal is grounded. The two coupled sidelines are connected to the output ports,
whereas an isolation resistor connects the other ends. The power division ratios can be tuned
by altering the capacitance of two centrally connected varactors. Adding capacitors between
coupled lines can affect the odd-mode impedance of the circuit and tuning the capacitance of
the varactors will further change the odd-mode impedance and phase velocity simultaneously.
There are two bias circuits applying voltages to two varactors; the bias circuits come with L
biases and C blocks connected to prevent short circuit. The results in Fig. 2. 12(a)-(c) show a
tunable power division ratio from 1:1 to 2.4:1 with more than 15 dB of isolation and 10 dB of
return loss across the frequency band from 0.7 to 1.4 GHz.
(a)
(b)
Fig. 2. 11 (a) Tri-coupled line structure, (b) layout and photography of the proposed
tunable PD [70].
P a g e | 30
(a)
(b)
(c)
Fig. 2. 12 Measured (a) S21 and S31, (b) S11 and the phase difference between output ports,
(c) S23 [70].
P a g e | 31
2. 2. 2 Tunable Operating Frequency
Similar to the tunable power division, from the reconfigurable operating frequency point of
view, the most commonly used method is adding varactors to change the ports impedance
matching. The tunability results from the impedance/phase velocity change of the coupled or
uncoupled lines.
Fig. 2. 13(a) (b) show the configuration and layout of the PD in [71]. Adapting a stepped-
impedance transformer on the common arm, a pair of coupled lines, and a shunt varactor is
connected at the common node for tunability. An approximate solution for impedance Z3 and
Z4e is given in the paper under the condition of θ3 is small, and the sum of θ3 and θ4 approaches
θ2. According to Fig. 2. 14(a)-(d), this work achieves a wide tuning band of 2.89:1 considering
all S-parameters better than 20-dB. This tuning band range is extensive in relation to its
compact size (0.24𝜆𝑔 × 0.02𝜆𝑔 ). However, as mentioned in the paper, the bandwidth is not
constant; with the change of centre frequency, the lowest return loss will turn up after 2 GHz,
resulting in the bandwidth shrink with a 20-dB standard. Although this study exhibits a very
wide tuning range, the main drawback is putting the isolation circuit between the closely spaced
coupled-line. Moreover, the realization of physical isolation between two outputs requires two
50Ω transmission lines which increases circuit size. Integrating filtering response into a PD as
shown in Fig. 2. 15(b) is effective for system simplification. For instance, a PD and tunable
filter combined in the same front-end can be merged as shown in Fig. 2. 15(a), resulting in a
multi-functional but simply component with no extra loss.
(a) (b)
Fig. 2. 13 (a) Configuration and (b) fabricated circuit of the PD.
P a g e | 32
(a) (b)
(c) (d)
Fig. 2. 14 Comparison of simulated and measured (a) S11, (b) S22, (c) S21, and (d) S32.
P a g e | 33
(a) (b)
Fig. 2. 15 (a) Cascaded a filter and a PD to form a filtering PD, and (b) multiple bandpass
filter to form tunable bandpass filter.
Research study [72] utilises the synthesis of coupled-resonator filters to produce a tunable PD
with filtering response. The art of this method is that tuning the varactors will only affect the
resonating frequencies, while the coupling coefficient and quality factors do not change. In
other words, the centre operating frequency will be changed by altering varactors, but the
bandwidth will be ideally kept constant. Since tunable filters with constant bandwidth have
become a popular trend, the same mechanism can be applied to design PDs.
P a g e | 34
(a) (b)
(c) (d)
Fig. 2. 16 (a) Configuration and (b) fabricated circuit of the proposed PD in [72], and
comparison of measured and simulated performances regarding (c) S11, and (d) S22 & S33.
The configuration of the PD is shown in Fig. 2. 16(a) (b). Methods of controlling the coupling
coefficients and external quality factors are investigated, and the constant bandwidth is
complimented by using capacitors as well as magnetic and electric coupling to obtain required
external quality factors and coupling coefficients. A constant bandwidth is maintained when
tuning the frequency. The performance is shown in Fig. 2. 16(c) (d) with the frequency tuning
range from 0.62 to 0.85 GHz (1.37:1) with an absolute bandwidth of 60±2.5 MHz. Since the
P a g e | 35
frequency tuning range is not very wide, this method is suitable for applications that require
the same absolute bandwidth in a small tuning range.
2. 2. 3 Hybrid Reconfigurable PDs
The word “hybrid” indicates multiple reconfigurable responses, including operating frequency
tuning, operating bandwidth tuning, and power division tuning, all implemented at the same
time.
Fig. 2. 17 Reconfigurable RF front end with tunable power dividing and controllable
filtering for antennas.
For example, in a reconfigurable RF front end, as shown in Fig. 2. 17, a tunable PD and two
tunable bandpass filters are usually cascaded. The tunable PD can control the power dividing
ratios for each output terminals, while the tunable filters can control the spectrum of the
transmitting signals to eliminate interferences. To simplify the system and avoid unnecessary
transmitting loss, hybrid reconfigurable PDs are very attractive. However, so far a
reconfigurable wideband filtering PD having a simultaneous frequency, bandwidth, and power
division control is less explored.
The design in [73] proposed a wideband in-phase PD with tunable power division ratio and
filtering response as depicted in Fig. 2. 18(a) (b). The design consists of a three-line coupled
structure loaded with a pair of varactors in between a pair of short-ended stubs at two outputs.
A variable power division is achieved by changing the coupling factors between the centre-line
and side-lines of the three-line coupled structure using a pair of varactors. The centre frequency
P a g e | 36
(a)
(b)
Fig. 2. 18 (a) From cascaded structure to an integrated tunable filtering PD, (b) proposed
design configuration, and (c) fabricated circuit (0.24𝝀𝒈 ×0.12𝝀𝒈 ) for [73].
and corresponding operating bandwidth are tuned by the varactors; one at the centre of the
loaded short-ended stub, and another one connecting the end of the stub to the ground. The
power division ratio is a function of capacitances. Once the mode impedance and loaded
varactors are defined, the power division ratio can be controlled easily by the two varactors
between the three coupled-line.
P a g e | 37
(a) (b)
(c) (d)
Fig. 2. 19 Simulations and measurements results for demonstrating (a) different power
division ratio, (b) bandwidth tunability, (c) centre frequency tunability, and (d) cut-off
edge tunability and in-band differential phase of [73].
To investigate the filtering response, the loaded varactors are kept at the same value. A series
of transmission zeros that located at harmonic frequencies 2f0, 4 f0 to 2n f0 (n = 1, 2,.) are
generated due to the three-line coupled structure. Similar to the two-line structure, reflection-
zeros result from the cancellation of phase at these frequencies. Moreover, more transmission
zeros can be produced by the shunted varactors loaded at the short-ended stubs. The bandwidth
and the centre frequency of the tunable filtering band highly depend on the position of
transmission zeros. In the experiment of [73], the bandwidth tuning range achieved 1.29:1 to
2.2:1 as shown in Fig. 2. 19(a)-(d), which is substantial. For the isolation circuit analysis, the
mutual coupling between the two side lines is extremely weak, and the equivalent capacitances
P a g e | 38
of the coupled lines do not vary significantly from their values at the centre frequency.
Therefore, the isolation remains almost the same under equal or unequal power division ratio.
To sum up, the proposed design has a controllable filtering bandwidth, relocatable cut-off zeros,
and tunable centre frequency. The Tunability of power division ratio ranges from 0.5:1 to 2:1;
controllable filtering bandwidth from 1.42:1 to 1.92:1; tunable centre frequency from 1.3 to
1.48 GHz; sharp cut-off with relocatable transmission zeros; and upper stopband harmonic
suppressions up to more than 6 GHz in the whole measurement process. The experimental
performance is slightly different from the simulated one due to the nonlinearity of the tuning
elements as well as losses in fabrication and measurements processes.
It is noted that the tunability of the centre frequency comes from the shunted varactor-loaded
stubs, while the resonant modes are related to both three-line coupled structure and shunted
stubs. This is the reason why the tunability of the centre frequency is limited to some extent.
This design can serve a range of narrowband to wideband applications where the tunability of
the bandwidth is more significant than that of the centre frequency.
Similar work using the same method is reported in [74]; the structure and tuning mechanism is
the same as shown in Fig. 2. 20(a) (b). The circuit consists of two tri-mode stub-loaded
resonators, a three-line coupled structure loaded two varactors, and two output parallel-coupled
lines loaded a lumped capacitor. The centre frequency and operating bandwidth of wideband
filtering response can be tuned by controlling the varactor diodes loaded onto the open end of
tri-mode resonators. Meanwhile, the power division ratio is controlled by tuning the coupling
strength between the centre line and two side-line of three-line coupled structure. The capacitor
loaded onto the parallel-coupled line of two output ports is to enlarge the coupling between
them, which can increase the return loss S11, and then can relieve the limitation of the line space
between two coupled lines. The advance of [74] over [73] is the wider centre frequency tuning
range, from 0.86 GHz to1.32 GHz. However, the occupied area for this design is slightly larger
than [73].
P a g e | 39
(a)
(b)
Fig. 2. 20 (a) Circuit model of reconfigurable wideband filtering PD and (b) fabricated
circuit for [74].
P a g e | 40
2.3 Coupled Resonator PDs
Resonant-coupled PDs exhibit a great prospect in communication systems due to the compact
size, harmonic suppression [75] [76], and high passband selectivity [77]. The quarter-
wavelength sections of a conventional Wilkinson PD can be replaced by coupled resonators to
reduce circuit size [78]. Furthermore, the critical resonant coupling condition between
resonators will only be achieved at certain frequencies, contributing the harmonics suppression.
Cross-coupling between the resonators will introduce transmission zeros. If the introduced
transmission zeros are located at the end of the passband, the transition between the passband
and the stopband will be sharpened. The selectivity of the passband can be enhanced by
increasing external coupling between resonators. Some design examples in terms of size
miniaturization and passband selectivity enhancement are discussed in this section. The work
in [79] has demonstrated a compact PD design with a filtering response using spiral resonators.
The spiral resonators are utilised for miniaturisation due to its curled structure. The coupling
scheme of the PD is shown in Fig. 2. 21(a).
(a) (b)
Fig. 2. 21 The (a) coupling scheme, and (b) the structure of the in-phase PD in [79].
The source-load cross-coupling is employed to create a pair of transmission zeros on both ends
of the passband. The PD is composed of four half-wavelength spiral resonators. Signals
transmitted from input will be split into two ways directly by the T-shape feedlines (represented
P a g e | 41
(a) (b)
(c) (d)
(e)
Fig. 2. 22 Selective band-edge regarding (a) L4 and (b) S4, and tuning frequency band
regarding (c) L1, and (d) W1, (e) comparison of simulated and measured results [79].
P a g e | 42
by the dash lines in Fig. 2. 21(a)). The feedlines will couple with the two pairs of spiral
resonators as shown in Fig. 2. 21(b). It is suggested in [79] that this type of resonator-coupling
scheme can provide strong operating-passband enhancement around the resonant frequency. In
addition to filtering response and high selectivity, the passband and the two transmission zeros
can be controlled by properly choosing the lengths and gaps of the resonators. A comprehensive
demonstration of the frequency responses with varied coupling stength can be seen in Fig. 2.
22(a)-(e). The transmission zeros can be controlled by changing the gaps between two
input/output feedlines, as shown in Fig. 2. 22 (a). The variation of the gap widths will change
the source-load cross-coupling conditions. Higher selectivity can be achieved by enhancing the
cross-coupling strength. The pictures in Fig. 2. 22(c) (d) illustrate that the centre frequency of
the passband can be tuned by changing the width and length of the spiral resonators. The
dimension of the resonators determines the resonant frequency so that leads to a change of
centre frequency. The measured results of the PD in Fig. 2. 22(e) show good agreement with
simulations. The fabricated circuit is a second order PD with a compact size of 0.15 λg×0.3 λg.
However, the isolation between the output ports is relatively poor due to the cross-coupling
between each port.
A compact and high in-band isolation resonator-based PD with a second-order Chebyshev

filtering response is proposed in [80]. The coupling scheme is shown in Fig. 2. 23(a). The size
can be reduced by modifying the structural frame of the quarter-wavelength unit-impedance
resonators. The traditional unit-impedance resonator shown in Fig. 2. 23(b) can be evolved into
the net-type resonator with two open-ended stubs and two shorted-ended stubs. For further
miniaturization, the net-type resonator is folded to a square box shape where the size is
approximately 0.09λg by 0.09λg. Moreover, the electric and magnetic coupling can be easily
achieved due to the configuration of the resonator. It should be noted that the coupling
coefficient 𝑀34 in Fig. 2. 23(a) is realised by the magnetic coupling, while the others are
realized by the electric coupling. Herein, the electric coupling can be realised when the open-
ended line sections of two folded net-type resonators are closely placed, and the magnetic
coupling can be achieved if the short-ended line sections of two folded net-type resonators are
closely allocated. The proposed coupling scheme is symmetric with respect to the two output
ports. When an incident signal is applied at output Port 2, the signal will be evenly split into
two out-of-phase parts. These two parts can cancel each other at output Port 3 and thus no
P a g e | 43
power is coupled to Port 3. This results in enhanced in-band isolation. The measurement results
in Fig. 2. 24(a) (b) depict a narrow bandwidth of input return loss with only 1.036:1 at the
centre frequency. The measured results show very good in-band isolation, which is better than
30 dB. It should be mentioned that the circuit has an extremely compact size of 0.19λg×0.19λg
which was achieved by curving quarter wavelength transmission lines into spiral resonators.
An in-band performance deterioration is obtained, which might be caused by fabrication and
measurement errors.
(a) (b)
(c)
Fig. 2. 23 (a) Coupling scheme of the PD in [80], and the (b) quarter-wavelength unit-
impedance resonator, and the (c) configuration of the PD in [80].
P a g e | 44
(a)
(b)
Fig. 2. 24 Comparison of simulation and measurement in terms of (a) |S11| and |S21|, and
(b) |S22|, |S33| and |S23|.
In general, resonator-based PDs exhibit remarkable features such as compact sizes, high
passband selectivity, and harmonic suppression. These features are realised by using the
coupling mechanism between the resonators. The majority of the resonator-based PDs have
P a g e | 45
these characteristics and as a result possess filtering responses. The resonator-based PD design
methods can be borrowed from filter designs. However, the isolation performance highly
depends on the coupling scheme.
2.4 Millimetre-Wave PDs
Millimetre-Wave (mm-Wave), as indicated by the name, has a frequency that ranges from 30
GHz to 300 GHz with such a small wavelength. Mm-Wave devices were exotic and only been
used as specialised applications such as military space. As low-frequency bands have been
assigned to numerous wireless communication systems and the number will subsequently keep
growing, high-frequency bands are becoming increasingly attractive and necessary. Mm-Wave
bandwidths have long been employed for military radar systems and are frequently being used
in commercial automotive collision-avoidance radar systems. Achieving mm-Wave circuit
designs on reliable Printed Circuit Boards (PCB) materials in a practical manner will be the
challenge in making these higher frequencies affordable.
Several Design Examples
1) Microstrip
A novel Monolithic Hybrid Microwave Integrated Circuit mm-Wave PD/combiner was

proposed in [81]. A ring PD/combiner was used, and it was integrated on a ceramic substrate.
The structure showed in Fig. 2. 25(a) enhances the traditional Wilkinson design by adding two
half-wave transmission lines to connect the integrated resistor, converting the layout in a ring
circuit of 1.5𝜆𝑔 length. These additional lines create the ideal framework for this isolation
resistor, reducing undesirable mutual coupling between the main transmission arms. The
measured transmission magnitudes that are less than 3.5 dB which can be seen from Fig. 2.
25(b), while the overall imbalance is less than 0.25 dB. Also, from Fig. 2. 25(c) the matching
of all ports is less than -15 dB, and the measured isolation factor between the two output ports
is at least 25 dB. Overall, this work shows an outstanding performance, but the bandwidth is
relatively narrow.
P a g e | 46
(a)
(b)
(c)
Fig. 2. 25 (a) Modified Wilkinson PD and micro-photographs of the PD in [81], and

measured results of (b) transmission, and (c) return losses and isolation.
P a g e | 47
2) Substrate Integrated Waveguide
Since the conventional microstrip structure will cause incredible energy loss in mm-Wave
frequencies, Substrate Integrated Waveguide (SIW) technology may be one of the solutions.
Traditional metallic waveguides exhibit excellent properties in designing high frequency, high
Q, and deficient loss devices. As modern communication systems require lightweight and
miniaturized components, one of the main drawbacks for metallic waveguide is the massive
size, which involves complex transitions to integrated planar circuits. The transitions always
consist of two or more separate pieces that need accurate assembly, and a tuning mechanism is
also generally essential. Furthermore, the planar substrate has to be cut into a specific shape.
These constraints make integration difficult and costly. The concept of integrated microstrip
and rectangular waveguide in planar form was firstly proposed by D. Deslandes and K. Wu
[82]. A basic SIW structure is shown in Fig. 2. 26: the rows of metallised via holes form the
electrical side walls, while the substrate is in between the top and bottom metal plates to allow
propagation of TE modes. This form of the waveguide will inevitably reduce the factor of the
waveguide due to dielectric filling and volume reduction. However, the entire circuit, including
planar circuit, transition, and waveguide can be constructed using standard PCB or other planar
processing techniques. In other words, SIW converts 3D bulky metallic waveguides to planar
ones and preserves the advantages of conventional waveguides.
Fig. 2. 26 Structure of SIW and its equivalent rectangular waveguide.
The SIW has recently become very popular for the design and development of high-
performance and cost-sensitive filters over a very large frequency range from sub-gigahertz to
sub-terahertz. Plenty of works [83]-[86] have focused on designing SIW filters with smaller
P a g e | 48
sizes and better performances. The mechanism of designing a PD is similar to a filter; energy
is transmitted to outputs within assigned frequency bands. Many SIW PDs [87]-[91] have been
proposed, but few of them concentrated on mm-Wave applications. The design in [92] reported
a broadband H-plane Wilkinson PD that makes use of the half-mode substrate integrated
waveguide (HMSIW) technique for the first time (see Fig. 2. 27(a)). This broadband PD utilizes
the half-mode structure of SIWs to achieve a half-size reduction. In addition, the lossy network
or resistor branch is integrated with the SIW structure in order to obtain good output matching
and isolation if unbalanced signals are combined or divided.
(a)
(b) (c)
Fig. 2. 27 (a) A prototype of the proposed HMSIW PD, and simulations and
measurements for (b) S11, S21 and S31, and (c) S22, S33 [92].
The design takes the advantages of SIW, such as low profile, low insertion loss, and little
interference, however the resulting structure becomes smaller in size. It should be noted that
the half-mode configuration does not support higher-order mode TE20 as in the case of SIW.
P a g e | 49
Therefore, a broader bandwidth is another side product of the proposed PD in contrast with its
counterparts. The proposed PD can easily be used at mm-Wave frequencies and favourable for
the microwave integrated circuit applications. Measured results in Fig. 2. 27(b) (c) suggest that
good input return loss S11 and S21 and S31 performances can be achieved across a broad
bandwidth from 18 GHz up to 40 GHz. Meanwhile, S22, S33, and S23 better than 10-dB are also
ensured across a 1.94:1 bandwidth from 18 GHz to 35 GHz. However, the design is critical for
power combining or non-equal power dividing due to the potentially poor output port matching.
3) Coplanar Waveguide
Several works [93]-[95] focus on the design of a conventional Wilkinson PD using coplanar
waveguides. The circuit consists of four quarter-wave coupled lines that are cleverly connected
in a ring configuration as shown in Fig. 2. 28.
(a)
(b)
Fig. 2. 28 (a) Simplified schematic of the coplanar ring divider [91], (b) Simulated surface
current flow at 80 GHz.
P a g e | 50
(a) (b)
(c) (d)
Fig. 2. 29 Measured results compared with Simulations of (a) S11, (b) S21, (c) S33 and (d)
S23.
The design is uniplanar and was fabricated on a thick GaAs substrate. The isolation resistor
was formed by using 33 nm thick of NiCr to give approximately 50 Ω/square. The air-bridge
connecting the ground planes of the CPW and the symmetric CPS crossover is used to prevent
odd-mode generation. Unlike the conventional Wilkinson divider, which has a pair of lumped
resistors, the proposed divider has two quarter-wave symmetric CPSs, a coplanar stripline
crossover, and a pair of 100 Ω resistors between the two output ports. The crossover forms a
phase inverter, which provides 180ºnon-frequency selective phase inversions; the two quarter-
wave CPSs ensure an open-circuit at the output port when looked into the other ports. The
divider is thus expected to have broadband isolation performance. The two resistors provide a
good port matching comparing to conventional Wilkinson PDs. The measurements in Fig. 2.
29(a)-(d) indicate a transmission loss lower than 1.3 dB and S11 better than 15 dB. The S23 is
P a g e | 51
(a)
(b) (c)
(d) (e)
Fig. 2. 30 (a) Structure of CPW ring PD and (b) S-parameters simulation and
measurement comparison of (b) S11, (c) S22, (d) SS21 and (e) S23 [95].
P a g e | 52
kept below 18-dB in the entire V and W bands. The performance is comparable to the devices
that used at lower frequencies, and the fabrication process is more straightforward compared
to studies at similar bands. Similar work has been conducted in [95]. A wideband CPW power
combiner is shown in Fig. 2. 30(a) using two upper quarter-wave CPWs and two lower quasi
quarter-wave CPWs with an impedance of 70.7 Ω. Compared to [93], this structure avoids high
insertion loss caused by CPW and CPS transitions.
Experimental results in Fig. 2. 30(b)-(e) indicate less than 0.8 dB transmission loss (less than
0.5 dB from simulation may cause by the transition of CPWs and ports). S11 and S22 are greater
than 15 dB, and S23 is better than 22 dB in the frequency range of 50 and 100 GHz. The results
have demonstrated that even in such high frequencies, efficient wideband power dividing and
combing can be achieved by using standard GaAs MMIC process and air-bridges.
2.5 Summary
This chapter has presented and discussed a large amount of state-of-the-are PDs from
conventional prototypes to multi-functional PDs. Different aspects of PD’s performance are
discussed in this chapter including operating bandwidth, power division and frequency
selectivity, etc. Normally, a variable bandwidth and power dividing ratio and broad frequency
selective range are desired. However, the circuits will have high complexity and larger size
when pursuing good performance. In this case, realizing excellent performance while
maintaining small electrical size and simply layout is a challenge for PD design. In addition to
performance, abundant functionality is the other remarkable challenge. Engineers always
facing specifications that require PD to have multi-function. In general, versatility is
implemented by cascading different components which sacrifices transmission efficiency.
Multi-functional PDs can integrate power dividing with filtering response. Besides, RF
components at mm-Wave frequencies have attracted more and more attention and can be
extensively investigated due to the promising applications in modern wireless communications.
Maintaining promising performance and avoiding unnecessary loss at mm-Wave frequencies
are also very challenging.
The work to be presented in Chapter 3 has remarkable operating bandwidths by using a

variable form of a conventional Wilkinson PD. It exhibits excellent performance with a simply
P a g e | 53
layout. Comparing to other similar work presented in this chapter, the proposed design can
achieve a wider bandwidth but the size is less than 2/3 of that of a conventional PD. The design
in Chapter 4 based on a Horst-type Wilkinson power divider can provide a great frequency
tuning range. Thus, high frequency selectivity and a simple structure are implemented at the
same time. The proposed design in Chapter 5 can be applied at mm-Wave frequencies.
Additionally, the proposed coupling strategy can provide filtering response and get rid of T- or
Y- junction at the common port for loss reduction. This design integrates good performance,
multi-function and low loss together at the same time. The proposed design method has a great
potential for the development of various mm-Wave components.
References
118, 1960.
[2] U. Gysel, “A new N-way power divider/combiner suitable for high power applications,” in 1975
MTsymp. Dig., pp.116-118.
[3] Y. Wu, Y. Liu, and X. Liu, “Dual-frequency power divider with isolation stubs,’’ Electronics
Letters, vol. 44, no. 24, pp. 1407-1408, 2008.
[4] Y. Wu, Y. Liu, Y. Zhang, J. Gao, and H. Zhou, “A dual band unequal Wilkinson power divider
without reactive components,’’ IEEE Transactions on Microwave Theory and Techniques, vol. 57,
no. 1, pp. 216-222, Jan. 2009.
[5] M. Bemani and S. Nikmehr, “Nonradiating arbitrary dual-band equal and unequal 1:4 series power
dividers based on CRLH-TL structures,’’ IEEE Transactions on Industrial Electronics, vol. 61, no.
3, pp. 1223-1234, Mar. 2014.
[6] N. Gao, G. Wu, and Q. Tang, “Design of a novel compact dual-band Wilkinson power divider with
wide frequency ratio,’’ IEEE Microwave and Wireless Components Letters, vol. 24, no. 2, pp. 81-
83, Feb. 2014.
[7] M. Liao, Y. Wu, Y. Liu, and J. Gao, “Impedance-transforming dual-band out-of-phase power
divider,’’ IEEE Microwave and Wireless Components Letters, vol. 24, no. 8, pp. 524-526, Aug.
2014.
[8] C. Cai, J. Wang, Y. Deng, and J.-L. Li, “Design of compact dual-mode dual-band filtering power
divider with high selectivity,’’ Electronics Letters, vol. 51, no. 22, pp. 1795-1796, 2015.
P a g e | 54
[9] X. Ren, K. Song, M. Fan, Y. Zhu, and B. Hu, “Compact dual-band Gysel power divider based on
composite right- and left-handed transmission lines,’’ IEEE Microwave and Wireless Components
Letters, vol. 25, no. 2, pp. 82-84, Feb. 2015.
[10] C.-W. Tang and Z.-Q. Hsieh, “Design of a planar dual-band power divider with arbitrary power
division and a wide isolated frequency band,’’ IEEE Transactions on Microwave Theory and
Techniques, vol. 64, no. 2, pp. 486-492, 2016.
[11] X. Wang, I. Sakagami, Z. Ma, A. Mase, M. Yoshikawa, and M. Ichimura, “Miniaturized dual-band
Wilkinson power divider with self-compensation structure,’’ IEEE Transactions on Components,
Packaging and Manufacturing Technology, vol. 5, no. 3, pp. 389-397, Mar. 2015.
[12] T. Zhang, W. Che, H. Chen, and W. Feng, “A compact four-way dual-band power divider using
lumped elements,’’ IEEE Microwave and Wireless Components Letters, vol. 25, no. 2, pp. 94-96,
Feb. 2015.
[13] W. Zhang, Z. Ning, Y. Wu, C. Yu, S. Li, and Y. Liu, “Dual-band out-of-phase power divider with
impedance transformation and wide frequency ratio,’’ IEEE Microwave and Wireless Components
Letters, vol. 25, no. 12, pp. 787-789, Dec. 2015.
[14] H. S. Gharehaghaji and H. Shamsi, ‘‘Design of unequal dual band Gysel power divider with
isolation bandwidth improvement,’’ IEEE Microw. Wireless Compon. Lett., vol. 27, no. 2, pp.
138–140, Feb. 2017.
[15] M.-J. Park, ‘‘Coupled line Gysel power divider for dual-band operation,’’ Electron. Lett., vol. 47,
no. 10, pp. 599–601, May 2011.
[16] Z. Sun, L. Zhang, Y. Liu, and X. Tong, ‘‘Modified Gysel power divider for dual-band
applications,’’ IEEE Microw. Wireless Compon. Lett., vol. 21, no. 1, pp. 16–18, Jan. 2011.
[17] M. A. Maktoomi and M. S. Hashmi, ‘‘A performance enhanced port extended dual-band
Wilkinson power divider,’’ IEEE Access, vol. 5, pp. 11832–11840, 2017.
[18] M. H. Maktoomi, D. Banerjee, M. S. Hashmi, "An enhanced frequency-ratio coupled-line dual-
frequency Wilkinson power divider", IEEE Trans. Circuits Syst. II Exp. Briefs. vol. 65, no. 7, pp.
888-892, July 2018.
[19] R. Gomez-Garcia, R. Loeches-Sanchez, D. Psychogiou, and D. Peroulis, “Single/multi-band
Wilkinson-type power dividers with embedded transversal filtering sections and application to
channelized filters,’’ IEEE Transactions on Circuits and Systems I: Regular Papers, vol. 62, no.
6, pp. 1518-1527, Jun. 2015.
[20] Q. X. Chu and F. Lin, “Design of multi-band microstrip power dividers,” in Proc. EDAPS, Dec.
2011, pp. 1–4.-
[21] Li, Y.C., Xue, Q., Zhang, X.Y.: ‘Single- and dual-band power dividers integrated with bandpass
filters’, IEEE Trans. Microw. Theory Techn., 2013, 61, (1), pp. 69–76
P a g e | 55
[22] K. Song, Y. Mo, Q. Xue, and Y. Fan, “Wideband four-way out-of-phase slotline power dividers,’
IEEE Transactions on Industrial Electronics, vol. 61, no. 7, pp. 3598-3606, Jul. 2014
[23] M. M. Honari, L. Mirzavand, R. Mirzavand, A. Abdipour, and P. Mousavi, “Theoretical design of
broadband multisection Wilkinson power dividers with arbitrary power split ratio,’’ IEEE
Transactions on Components, Packaging and Manufacturing Technology, vol. 6, no. 4, pp. 605-
612, 2016.
[24] W. Feng, H. Zhu, W. Che, and Q. Xue, “Wideband in-phase and out-of-phase balanced power
dividing and combining networks,’’ IEEE Transactions on Microwave Theory and Techniques,
vol. 62, no. 5, pp. 1192-1202, May 2014.
[25] M. Strackx, K. Janssen, E. D’Agostino, G. Vandenbosch, P. Reynaert, and P. Leroux, “Ultra-
wideband antipodal Vivaldi antenna array with Wilkinson power divider feeding network,” in Proc.
IEEE ICUWB, 2011, pp. 1–4.
[26] Liu, W.-Q., Wei, F., Pang, C.-H., and Shi, X.-W.: ‘Design of a compact ultra-wideband power
divider’. Int. Conf. Microwave Millimeterwave Tech. (ICMMT), Shenzhen, China, May 2012, pp.
1–3
[27] Zhu, H., Abbosh, A., and Guo, L.: ‘Wideband four-way filtering power divider with sharp
selectivity and wide stopband using looped coupledline structures’, Microw. Wirel. Compon. Lett.,
2016, 26, (6), pp. 413–415.
[28] U. Ahmed and A. Abbosh. “Compact power divider for wideband in-phase and out-of-phase
performances using parallel coupled lines,” Electronic Letters, vol: 53, issue: 19. Pp.1312-1314.
Spe, 2017.
[29] Packiaraj, D., M. Ramesh, and A. T. Kalghatgi. "Broadband Equal Power Divider." Journal of
Microwaves, Optoelectronics and Electromagnetic Applications 16, no. 2 (2017): pp. 363-370.
[30] Y. Wu, L. Jiao, Z. Zhuang, and Y. Liu, “The art of power dividing: A review for state-of-the-art
planar power dividers,” China Communications, vol. 14, no. 5, pp.1-16, May 2017.
[31] S. B. Cohn, “A class of broadband three-port TEM-mode hybrids,” IEEE Trans. Microw. Theory
Techn., vol. MTT-16, no. 2, pp. 110-116, Feb. 1968.
[32] J.-Q. Gong, K. Gao, and C.-H. Liang, “Synthesis of a miniaturised wireless local area network
dual-band microstrip Wilkinson power divider through a general three-step optimisation process,”
IET Microw., Antennas Propag., vol. 9, no. 12, pp. 1274–1278, Sep. 2015.
[33] L. Young, "Tables for cascaded homogeneous quarter-wave transformers (correction)", IRE Trans.
Microw. Theory Tech., vol. MTT-8, no. 3, pp. 243-244, Mar. 1960.
[34] S. W. Lee, C. S. Kim, K. S. Choi, J. S. Park, D. Ahn, "A general design formula of multi-section
power divider based on singly terminated filter design theory", IEEE MTT-S Int. Microw. Symp.
Dig., vol. 2, pp. 1297-1300, 2001.
P a g e | 56
[35] T, Yu, “A broadband Wilkinson power divider based on the Segmented Structure”, IEEE Trans.
Microw. Theory Techn., vol:66, Issue: 4, pp. 1902-1911, Feb. 2018.
[36] L. Young, "Stepped-impedance transformer and filter prototypes," IRE Trans. Microwave Theory
Tech., vol. MTT-10, pp. 339-359, Sept. 1962.
[37] Hong, J.S.G. and Lancaster, M.J., 2004. Microstrip filters for RF/microwave applications (Vol.
167). John Wiley & Sons.
[38] S.-H. Chien & Y.-S. Lin. “Novel Wideband Absorptive Bandstop Filters With Good Selectivity”
IEEE Access, 2017. Vol 5, pp. 18847-18861.
[39] M. M. Nizam, P. J. Soh and A.H.M Suhaizal “"Design of a Modified L-shaped Bandstop Filter for
UWB applications." In Microwave and Optoelectronics Conference (IMOC), 2009 SBMO/IEEE
MTT-S International, pp. 177-182. IEEE, 2009.
[40] F.-C. Chen, R.-S. Li, J.-M. Qiu, and Q.-X. Chu, ‘‘Sharp-rejection wideband bandstop filter using
stepped impedance resonators,’’ IEEE Trans. Compon., Packag., Manuf. Technol., vol. 7, no. 3,
pp. 444–449, Mar. 2017.
[41] S. W. Lee, C. S. Kim, K. S. Choi, J. S. Park, D. Ahn, "A general design formula of multi-section
power divider based on singly terminated filter design theory", IEEE MTT-S Int. Microw. Symp.
Dig., vol. 2, pp. 1297-1300, 2001.
[42] A. M. Abbosh, "A compact UWB three-way power divider", IEEE Microw. Wireless Compon.
Lett., vol. 17, no. 8, pp. 598-600, Aug. 2007.
[43] H. Y. Yee, F.-C. Chang, and N. F. Audeh, “ N-way TEM-mode broadband power dividers,” IEEE
Trans. Microw. Theory Tech., vol. MTT-18, no. 10, pp. 682–688, Oct. 1970.
[44] J. Zhou and K. Morris, “Effects of interconnecting transmission lines on four-way Wilkinson
power divider,” Microw. Opt. Technol. Lett., vol. 51, no. 12, pp. 2850–2852, 2009.
[45] J. Zhou, K. A. Morris, and M. J. Lancaster, “General design of multi-way multi-section power
dividers by interconnecting two-way dividers,” IEEE Trans. Microw. Theory Tech., vol. 55, no.
10, pp. 2208–2215, Oct. 2007.
[46] Y. Xu and R. G. Bosisio, “Design of multiway power divider by using stepped-impedance
transformers,” IEEE Trans. Microw. Theory Tech., vol. 60, no. 9, pp. 2781–2790, Sep. 2012.
[47] L. Young, “Tables for cascaded homogeneous quarter-wave transformers,” IRE Trans. Microw.
Theory Tech., vol. MTT-7, no. 3, pp. 233–237, Apr. 1959.
[48] L. Young, “Tables for cascaded homogeneous quarter-wave transformers (correction),” IRE Trans.
Microw. Theory Tech., vol. MTT-8, no. 3, pp. 243–244, Mar. 1960.
[49] G. L. Matthaei, L. Young, and E. M. T. Jones, Microwave Filters, Impedance Matching Networks
and Coupling Structures. New York: McGraw-Hill, 1980.
P a g e | 57
[50] W.-J. Feng, M.-L. Hong, W.-Q. Che, Q. Xue. “Dual-Band Microstrip Bandstop Filter With
Multiple Transmission Poles Using Coupled Lines” IEEE Microwave and Wireless Components
Letters. Pp. 236-238, Vol: 27, Issue: 3, March 2017.
[51] W. Wang, M. Liao, Y. Wu, Y. Liu.” Small-size high-selectivity bandstop filter with coupled-line
stubs for dual-band applications.” Electronics Letters [serial on the Internet]. (2014, Feb 13), 50(4):
286-288
[52] J. Hong, “Microstrip filters for RF/microwave applications.” [electronic book]. Hoboken, N.J.:
Wiley, 2011.
[53] J. Hong & J. M. Lancaster, “Couplings of microstrip square open-loop resonators for cross-coupled
planar microwave filters.” Microwave Theory and Techniques, IEEE Transactions on, 44(11),
(1996): 2099-2109
[54] Y. Luo,& Q. Chu. “A Compact High Selectivity Dual-Band Bandstop Filter Using Bent L-
Resonators.” Microwave Conference (EuMC), 2013 European.
[55] H. Lee, C. Kim, W. Lim. “Compact wideband bandstop filters with electrically coupled open stubs.”
Microwave And Optical Technology Letters [serial on the Internet]. (2013, Nov 1), 55(11): 2701-
2703.
[56] K. Song, Y. Zhu, Q. Duan, M. Fan, and Y. Fan, “Extremely compact ultra-wideband power divider
using hybrid slotline/microstrip-line transition,’’ Electronics Letters, vol. 51, no. 24, pp. 2014-
2015, 2015.
[57] K. Song, Y. Mo, Q. Xue, and Y. Fan, “Wideband four-way out-of-phase slotline power dividers,’’
IEEE Transactions on Industrial Electronics, vol. 61, no. 7, pp. 3598-3606, Jul. 2014.
[58] U. T. Ahmed and A. M. Abbosh, “Wideband out-of-phase power divider using tightly coupled
lines and microstrip to slotline transitions,’’ Electronics Letters, vol. 52, no. 2, pp. 126-128, 2016.
[59] K. Song, Y. Zhu, Q. Duan, M. Fan, and Y. Fan, “Extremely compact ultra-wideband power divider
using hybrid slotline/microstrip-line transition,’’ Electronics Letters, vol. 51, no. 24, pp. 2014-
2015, 2015.
[60] S. Avrillon, I. Pele, A. Chousseaud, and S. Toutain, “Dual-band power divider based on semiloop
stepped-impedance resonators,” IEEE Trans. Microw. Theory Tech., vol. 51, no. 4, pp. 1269–1273,
Apr.2003.
[61] C.Wang and V. F. Fusco, “Passive broadband millimeter-wave uniplanar MMIC balun,” IEEE
Electronics Lett., vol. 44, no. 21, October 2008.
[62] L. Wu, Z. Sun, H. Yilmaz and M. Berroth, “A dual-frequency Wilkinson power divider,” IEEE
Trans. Microwave Theory Tech., vol.56, no. 1, pp. 278-284, January 2000.
P a g e | 58
[63] L. Chiu and Q. Xue, “A parallel strip ring power divider with high isolation and arbitrary power-
dividing ratio,” IEEE Trans. Microwave Theory Tech., vol.55, no. 11, pp. 2419-2462, November
2000.
[64] W. Feng, X. Ma, W. Che and Y. Wu. “Narrow-band balanced filtering network using coupled lines
loaded with stubs,” Electronics Letters. Vol: 54, issue: 6, pp. 366-368. March, 2018.
[65] Y. Wu and Y. Liu, “A novel wideband coupled-line Gysel power divider with function of
impedance matching,’’ Journal of Electromagnetic Waves and Applications, vol. 26, no. 14- 15,
pp. 2012-2021, Oct. 2012.
[66] R. B. Ekinge, “A new method of synthesizing matched broadband TEM mode three-ports,” IEEE
Trans. Microw. Theory Tech., vol. MTT-19, no. 1, pp. 81–88, Jan. 1971.
[67] X. Tang and K. Mouthaan, “Analysis and design of compact two-way Wilkinson power dividers
using coupled lines,” in Asia–Pacific Microw. Conf., Dec. 7–10, 2009, pp. 1319–1322.
[68] Wong, S.W., and Zhu, L.: ‘Ultra-wideband power divider with good in-band splitting and isolation
performances’, IEEE Microw. Wirel. Compon. Lett., 2008, 18, (8), pp. 518–520.
[69] C.-W. Tang and J.-T. Chen, ‘‘A design of 3-dB wideband microstrip power divider with an ultra-
wide isolated frequency band,’’ IEEE Trans. Microw. Theory Techn., vol. 64, no. 6, pp. 1806–
1811, Jun. 2016.
[70] L. Guo, H. Zhu, and A. M. Abbosh, “Wideband tunable in-phase power divider using three-line
coupled structure,” IEEE Microw. Wireless Compon. Lett., vol. 26, no. 6, pp. 404–406, Jun. 2016.
[71] T. Zhang and W. Che, “Compact tunable power divider with wide tuning frequency range and
good reconfigurable responses,’’ IEEE Transactions on Circuits and Systems II: Express Briefs,
pp. 1-1, 2016.
[72] L. Gao, X. Y. Zhang, and Q. Xue, “Compact tunable filtering power divider with constant absolute
bandwidth,’’ IEEE Transactions on Microwave Theory and Techniques, vol. 63, no. 10, pp. 3505-
3513, Oct. 2015.
[73] H. Zhu, A M. Abbosh, and L. Gao, “Planar In-Phase Filtering Power Divider With Tunable Power
Division and Controllable Band for Wireless Communication Systems,” IEEE Transactions on
Components, Packaging and Manufacturing Technology, vol. 8, no. 8, pp. 1458-1468, Aug. 2018.
[74] C. Zhu, J. Xu and W. Kang, “Microstrip Multifunctional Reconfigurable Wideband Filtering
Power Divider with Tunable Center Frequency, Bandwidth, and Power Division,” IEEE
Transactions on Microwave Theory and Techniques, VOL. 66, NO. 6, JUNE 2018
[75] Y. Wang, X. Y. Zhang, F. X. Liu and J. C. Lee, ”A compact bandpass wilkinson power divider
with ultra-wide band harmonic suppression,” IEEE Microw. Wireless Compon. Lett., vol. 27, no.
10, pp. 888-890, 2017.
P a g e | 59
[76] Gao, S.S., Sun, S., and Xiao, S.: ‘A novel wideband bandpass power divider with harmonic-
suppressed ring resonator’, Microw. Wirel. Compon. Lett., 2013, 23, (3), pp. 119–121.
[77] Zhang, B., and Liu, Y.: ‘Wideband filtering power divider with high selectivity’, Electron. Lett.,
2015, 51, (23), pp. 1950–1952.
[78] C.-F. Chen, T.-Y. Huang, T.-M. Shen, and R.-B. Wu, “Design of miniaturized filtering power
dividers for system-in-a-package,” IEEE Trans. Comp., Packag., Manufact. Technol., vol. 3, no.
10, pp. 1663–1672, Oct. 2013.
[79] K. Song, X. Ren, F. Chen, and Y. Fan, “Compact in-phase power divider integrated filtering
response using spiral resonator,” IET Microw. Antennas Propag., vol. 8, no. 4, pp. 228–234, 2014.
[80] Chen, C.F., and Lin, C.Y.: ‘Compact microstrip filtering power dividers with good in-band
isolation performance’, IEEE Microw. Wirel. Compon. Lett., 2014, 24, (1), pp. 17–19.
[81] D. Hammou, E. Moldovan, and S. O. Tatu, “Novel MHMIC millimeter wave power
divider/combiner,” in Proc. CCECE, Niagara Falls, ON, Canada, May 2011, pp. 280–283.
[82] D. Deslandes and K. Wu, "Integrated Microstrip and Rectangular Waveguide in Planar Form",
Microwave and Wireless Components Letters, IEEE, vol. 11, no. 2, Feb. 2001, pp 68 -70.
[83] Zhang, Q.-L., Yin, W.-Y., He, S., and Wu, L.-S.: ‘Compact substrate integrated waveguide (SIW)
bandpass filter with complementary split-ring resonators (CSRRs)’, IEEE Microw. Wirel. Compon.
Lett., 2010, 20, (8), pp. 426–428, doi: 10.1109/ LMWC.2010.2049258
[84] F. Huang, J.Y. Zhou, W. Hong, ‘Ku band continuously tunable circular cavity SIW filter with one
parameter,’ IEEE Microw. Wirel. Compon. Lett., 26 (4) (2016), pp. 270-272.
[85] M. Li, C. Chen, and W. Chen, “Miniaturized dual-band filter using dual-capacitively loaded SIW
cavities,” IEEE Microw. Wireless Compon. Lett., vol. 27, no. 4, pp. 344–346, Apr. 2017.
[86] X. Xu, J. P. Wang, and L. Zhu, “A new approach to design differentialmode bandpass filters on
SIW structure,” IEEE Microw. Wireless Compon. Lett., vol. 23, no. 12, pp. 635–637, Dec. 2013.
[87] H. Chen, W. Che, X. Wang and W. Feng, "Size-Reduced Planar and Nonplanar SIW Gysel Power
Divider Based on Low Temperature Cofired Ceramic Technology," in IEEE Microwave and
Wireless Components Letters, vol. 27, no. 12, pp. 1065-1067, Dec. 2017.
[88] J. Wang, J. Ni, Y.-X. Guo, and D. Fang, “Miniaturized microstrip Wilkinson power divider with
harmonic suppression,” IEEE Microw. Wireless Compon. Lett., vol. 19, no. 7, pp. 440–442, Jul.
2009.
[89] H. Chen, W. Che, Y. Cao, W. Feng, and K. Sarabandi, “Functionreconfigurable between SPDT
switch and power divider based on switchable HMSIW unit,” IEEE Microw. Wireless Compon.
Lett., vol. 27, no. 3, pp. 275–277, Mar. 2017.
P a g e | 60
[90] X. Wang and X.-W. Zhu, ‘‘Quarter-mode circular cavity substrate integrated waveguide filtering
power divider with via-holes perturbation,’’ Electron. Lett., vol. 53, no. 12, pp. 791–793, Aug.
2017.
[91] Danaeian, M., Moznebi, A.-R., Afrooz, K., and Hakimi, H.: ‘Miniaturized equal/unequal SIW
power divider with bandpass response loaded by CSSRs’, Electron. Lett., 2016, 52, (10), pp. 1864–
1866.
[92] Z.-Z. Yu and K. Wu. “Broadband half-mode substrate integrated waveguide (HMSIW) Wilkinson
power divider.” 2008 IEEE MTT-S International Microwave Symposium Digest (2008): 879-882.
[93] C. Li, , L. B. Lok, A. Khalid, V. Papageorgiou, J. Grant, and D. R. S. Cumming, “A coplanar ring
power divider with high isolation for Vband and W-band applications,” Proceedings of the 42nd
European Microwave Conference, 57–60, 2012.
[94] C. Li, L. B. Lok, A. Khalid, and D. R. S. Cumming, “Coplanar ring divider with wideband high
isolation performance,” Progress Electromagn. Res. Lett., vol. 25, pp. 1–10, 2011.
[95] C. Li, A. Khalid, V. Papageorgiou, L. B. Lok, and D. R. S. Cumming, “A Wideband CPW Ring
Power Combiner With Low Insertion Loss and High Port Isolation,” IEEE Microw. Wirel. Compon.
Lett., vol. 22, no 11, pp. 580-582, November 2012.
P a g e | 61
Chapter 3: Design of a Broadband Wilkinson Power Divider with Wide Range Tunable Bandwidths
Chapter 3 Design of a Broadband Wilkinson Power

Divider with Wide Range Tunable Bandwidths
This chapter presents a novel two-way and a cascaded eight-way power divider with a wide
range of tunable bandwidths. The proposed two-way power divider is designed based on adding
a pair of capacitors in parallel with the transmission lines in a Trantanella-type of Wilkinson PD.
This PD has an additional zero-reflection frequency which will significantly increase the
bandwidth compared with a conventional Wilkinson PD. Moreover, the added zero- reflection
frequency is sensitive to the capacitances so that the bandwidth can be tuned by tuning the
capacitors. Analytic design methods and formulas are derived and presented in detail. A two-
way PD is designed and measured to validate the proposed method. The measurement indicates
that the device can achieve a 15 dB bandwidth of 𝒇𝑯 /𝒇𝑳 =2.73:1 (0.75 GHz to 2.05 GHz). The
proposed method can be applied for the design of multi-way multi-section PDs. Meanwhile, an
eight-way PD is designed and fabricated. The measured response has a 13 dB bandwidth of
𝒇𝑯 /𝒇𝑳 =2.67:1 (0.75 GHz to 2 GHz). The proposed PDs exhibit wide bandwidths, compact sizes,
and good physical and electrical isolation features. The measurement agrees well with the
theoretical prediction which validates the proposed design theory.
3.1 Introduction
The features of PDs regarding bandwidth, circuit size, insertion loss, and isolation will
significantly influence the overall performance of the communication system. A Wilkinson-
type PD [1] is one of the most widely used dividers due to its simple structure and low insertion
loss. However, there are several drawbacks for the conventional Wilkinson PD. One is the
narrow bandwidth, and the other weakness is the lack of physical isolation for the two outputs.
In 2010, Trantanella [2] presented a novel PD by placing the isolation elements between the
λ/4 transmission lines at an arbitrary phase angle instead of 90° as in a conventional Wilkinson
divider to enhance both electrical isolation and physical separation. The structure has been
investigated in depth in [3]. Although the layout of the Trantanella Wilkinson PD has been
simplified, the bandwidth is limited to 1.5:1. Since there has been a sustained increase in the
P a g e | 62
demand of the wideband PDs, researches on broadband PDs have been proposed [4]-[6] to
increase the effective bandwidth. However, most of them are designed based on quarter-wave
transmission lines so that they suffer from the lack of compactness. Large sizes will make the
integration very challenging in some wireless devices. Nowadays, frequency-tunable devices
are attracting more and more attention because they can integrate diverse operating frequency
bands in a single device [4] [5] to satisfy multi-standard systems. A reconfigurable PD based
on the Trantanella structure was presented in [6] where the centre frequency can be changed
from 0.85 GHz to 2.4 GHz by tuning the bias voltage on the varactor diodes. However, the
effective bandwidth for each operational band is narrow. A few works [4]-[6] have been
proposed to provide tunable centre frequency and tunable bandwidth features within one
component, where the 10 dB bandwidth can achieve 2.32:1. These reconfigurable PDs have a
high potential to be widely used in multi-standard systems. But unfortunately, the complexity
of the circuit has been greatly increased, and also the10 dB in-band performance is normally
not sufficient to many systems.
A new broadband PD with a wide range of tunable bandwidths is presented in this chapter. The
idea is to add a pair of capacitors in parallel with the transmission lines as shown in Fig. 3. 1.
The proposed structure introduces an additional zero-reflection frequency to broaden the
operational bandwidth and make the bandwidth tunable. The idea of inserting a parallel
capacitor has been reported in [7] to improve the out-of-band selectivity. The relationship
between the variations of capacitance and the return loss has not yet been analytically discussed.
This chapter provides mathematic proof and design guidance on controlling the added zero-
reflection frequency for tunable bandwidths. The structure and design theory are discussed in
Section 3.2. The relationships between the electrical length of transmission lines, zero-
reflection frequency, and capacitance of added capacitors are illustrated and analysed in
Section 3.3. In Section 3.4, a two-way and an eight-way three-section PD are designed,
fabricated and measured to validate the proposed design method. Conclusions are drawn finally
in Section 3.5.
P a g e | 63
3.2 Design of the PD
A conventional Wilkinson PD is perfectly matched at all ports at the centre frequency so that
there is only one reflection-zero in the passband. To broaden the bandwidth, an additional
reflection-zero at 𝑓𝑧 can be generated by adding a pair of capacitors. As shown in Fig. 3. 1. 𝐶𝑃
denotes the added capacitors. 𝑍0 stands for the port impedance which is 50 Ω here. 𝜃1 , 𝜃2 , and
𝜃3 represent the electrical lengths of the transmission lines at the initial centre frequency, which
is assumed to be 1 GHz here. 𝑍1 , 𝑍2 , and 𝑍3 are the characteristic impedances of the
transmission lines. 𝑍𝐼𝑆𝑂 represents the complex impedance isolation component. There will be
two minimum-reflection frequencies for the proposed divider. One is near the original zero-
reflection frequency 𝑓𝑜 = 1 GHz and the other is the additional zero-reflection frequency 𝑓𝑧 .
By carefully choosing the values of the capacitors, the additional zero-reflection frequency 𝑓𝑧
can significantly extend the operational bandwidth of the device. Details will be discussed in
the following subsections.
Fig. 3. 1 The proposed broadband bandwidth tunable PD.
P a g e | 64
Return Loss
Fig. 3. 2 Even-mode equivalent circuit for the proposed divider.
The equivalent circuit of the proposed PD under even-mode excitation is shown in Fig. 3. 2.
The circuit can be divided into three sub-networks. Network 1 in the even-mode circuit contains
one front transmission and a capacitor that are connected to the front transmission line in parallel.
Network 2 is a single transmission line, and Network 3 is an open stub in the isolation branch.
To analyze the even-mode equivalent circuit, ABCD parameters for the first network need to be
calculated. Firstly, the normalized Y matrix of capacitor 𝐶𝑝 and the front transmission line can
be derived:
̅̅̅̅ 𝑗𝑌̅ −𝑗𝑌̅

|𝑌 𝐶𝑝 | = | | (3.1)
−𝑗𝑌̅ 𝑗𝑌̅
where 𝑌̅ = 2𝜋𝑓 ̅ ∙ 𝐶𝑃̅ is the normalized admittance of the capacitor 𝐶𝑃 , and 𝑓 ̅ is an arbitrary
frequency. The normalized Y matrix for the front transmission line is
𝑗 𝑗
− ̅̅̅
𝑍1 ∙tan(𝜃1 𝑓̅) 𝑍1 ∙sin(𝜃1 𝑓̅)
̅̅̅
|Y𝑇𝑥1 | = | 𝑗 𝑗 | (3.2)
− ̅̅̅
1 ∙sin(𝜃1 𝑓̅)
̅̅̅
𝑍 𝑍1 ∙tan(𝜃1 𝑓̅)
The ABCD matrix of the Network 1 can be derived by adding the Y matrixes in (1) and (2) and
then convert the sum to the ABCD matrix.
P a g e | 65
𝑗 𝑗
𝑗𝑌̅ − ̅̅̅ −𝑗𝑌̅ + ̅̅̅
𝑍1 ∙tan(𝜃1 𝑓̅) 𝑍1 ∙sin(𝜃1 𝑓̅)
̅̅̅̅
|𝑌 ̅̅̅̅̅̅
𝐶𝑝 | + |Y𝑇𝑥1 | = | | (3.3)
𝑗 𝑗
−𝑗𝑌̅ + ̅̅̅ 𝑗𝑌̅ −
𝑍1 ∙sin(𝜃1 𝑓̅) 𝑍1 ∙tan(𝜃1 𝑓̅)
̅̅̅
̅− 𝑗
𝑗𝑌 ̅̅̅̅ ̅)
𝑍1∙tan(𝜃1𝑓 1
̅− 𝑗 ̅− 𝑗
| 𝑗𝑌 ̅̅̅̅ ̅)
𝑍1∙sin(𝜃1𝑓
𝑗𝑌 ̅̅̅̅ ̅)
|
̅ 𝐵̅ |
|𝐴 = 2 2 (3.4)
𝐶 ̅ ̅ network1
𝐷 ̅ 𝑗 ̅− 𝑗 ̅− 𝑗
|(𝑗𝑌−̅̅̅̅ ̅ ))
𝑍1∙tan(𝜃1𝑓
−(𝑗𝑌 ̅̅̅̅ ̅ ))
𝑗𝑌 ̅̅̅̅ ̅ )|
𝑍1∙tan(𝜃1𝑓
̅+ 𝑗 ̅− 𝑗
𝑗𝑌 ̅̅̅̅ ̅) 𝑗𝑌 ̅̅̅̅ ̅)
𝑍1∙sin(𝜃1𝑓 𝑍1∙sin(𝜃1𝑓
Then the ABCD matrices for Network 2 and 3 can be expresses as:
̅ cos(𝜃2 𝑓 )̅ 𝑗𝑍2 sin(𝜃2 𝑓)̅

|𝐴 𝐵̅ | = | 𝑗 sin(𝜃2 𝑓̅) | (3.5)
𝐶̅ ̅ network2
𝐷 cos(𝜃2 𝑓 )̅
𝑍2
̅ 1 0
|𝐴 𝐵̅ | = |𝑗 tan(𝜃3 𝑓̅) 1| (3.6)
𝐶̅ ̅ network3
𝐷 𝑍 3
According to the sequence of these three networks, the ABCD matrix of the circuit can be
obtained by multiplying the ABCD matrices of the networks:
̅ 𝐵̅ | = |𝐴̅ 𝐵̅ | ̅ 𝐵̅ | ̅ 𝐵̅ |
|𝐴 ∙ |𝐴 ∙ |𝐴 (3.7)
𝐶 ̅ ̅
𝐷 𝐶̅ ̅
𝐷 network1 𝐶 ̅ ̅
𝐷 network3 𝐶 ̅ ̅ network2
𝐷
Hence:
A B

C D
1
Y 1
2 tan(1 f )
1
1 j (Y - )
Y 2 sin( 1 f )
2 sin( 1 f )
1
1 1 Y
（Y - ）2
- (Y  )2 2 tan(1 f )
2 sin( 1 f ) 2 tan(1 f )
1
1 Y-
j (Y - ) 2 sin( 1 f )
2 sin( 1 f )
1 0 cos( 2 f ) j 2 sin(  2 f )
 tan( 3 f )  sin(  2 f )
j j
2 1 2 cos( 2 f )
(3.8)
P a g e | 66
where the characteristic impedance of the transmission lines are normalized to 50 Ω, and the
frequency is normalized to the original centre frequency 𝑓𝑜 = 1 GHz . To simplify the
̅̅̅1 = 𝑍
calculation, the transmission lines have equal normalized characteristic impedances𝑍 ̅̅̅2 =
𝑍3 = √2. 𝜃𝑖 𝑓 ̅ (where i=1, 2, 3) represents the electrical length of each transmission line at
̅̅̅
normalized frequency 𝑓.̅ The normalized input impedance of Port 1 at 𝑓𝑧 should satisfy:
A B
Z inev   2  j0
CD (3.9)
̅̅̅̅̅
The real part of the input impedance should satisfy Re(𝑍 𝑒𝑣
𝑖𝑛 ) = 2. The capacitance 𝐶𝑃 can be
obtained by:
tan( 3 f Z )(tan(1 f Z ) tan( 2 f Z )  1)

Y  2 f Z  CP 
2
2(tan(1 f Z ) tan( 3 f Z )   2)
cos(1 f Z )
(3.10)
where 𝑓̅𝑧 is the normalized zero-reflection frequency. The imaginary part of the input
̅̅̅̅̅
impedance should satisfy Im(𝑍 𝑒𝑣 ̅
𝑖𝑛 ) = 0. The relationship between 𝜃𝑖 (i=1, 2, 3) and 𝑓𝑧 can be
expressed as:
tan( 3 f Z ) tan(1 f Z )  1
 2 tan( 3 f Z )  tan(1 f Z )
tan( 2 f Z ) tan( 2 f Z ) tan(1 f Z )  1

tan(1 f Z ) 4 2
 2 tan( 3 f Z ) tan(1 f Z )  4  2
tan( 2 f Z ) cos(1 f Z） cos(1 f Z )
 tan(1 f Z )
tan( 3 f Z )
(3.11)
To simplify the equation, the formulas can assume some restrictive conditions. For example, for
Trantanella dividers, there is usually the condition of 𝜃2 = 90° − 𝜃1 . For practical applications,
the electrical length of the additional transmission lines 𝜃3 could be any value but is usually
P a g e | 67
chosen between 0° to 15° referring to [3]. In this chapter, 𝜃3 is 7° . For any chosen zero-
reflection frequency 𝑓𝑧̅ (1 < 𝑓𝑧̅ < 3 ), 𝜃1 can be obtained from (3.11), given that 𝜃2 = π⁄2 −
𝜃1 and 𝜃3 is a fixed value. Now the value of 𝐶𝑃 can be calculated by (3.10).
Isolation Circuit
Fig. 3. 3 Odd-mode equivalent circuit for the proposed divider.
The performance of isolation can be obtained by using the odd-mode analysis method. The
theoretical values of components in the isolation circuit 𝑍𝐼𝑆𝑂 can be found under the condition
of perfect matching for port 2 and port 3 at the centre frequency 𝑓𝑐 = (𝑓0 + 𝑓𝑧 )/2. Fig. 3. 3
̅̅̅̅̅
shows the odd-mode equivalent circuit of the proposed PD.𝑍 𝑜𝑑 ̅̅̅̅̅
𝑜𝑑 ̅̅̅̅̅̅
𝑜𝑑 ̅̅̅̅̅̅̅
𝑜𝑑
1 , 𝑍2 , 𝑍3 , and 𝑍𝑡𝑜𝑡 are the
normalized output impedances of the circuit respectively, and they can be obtained by using:
1
j 2 tan(1 f c ) 
jY
Z1od 
1
j 2 tan(1 f c ) 
jY (3.12)
Z ISO
 j 2 tan( 3 f c )
Z 2od  2 2
Z
2  j ISO tan( 3 f c )
2 (3.13)
P a g e | 68
Z1od  Z 2od
Z od
3 
Z1od  Z 2od (3.14)
Z 3od  j 2 tan( 2 f c )
Z od
 2
2  j Z 3od tan( 2 f c )
tot
(3.15)
2 2 (tan( 1 f c )  tan( 3 f c )  tan(1 f c ) tan( 2 f c ) tan( 3 f c )  2 Y tan(1 f c ) tan( 3 f c ))
Z ISO  
2 (tan( 1 f c ) tan( 2 f c ) tan( 3 f c )  2 Y tan(1 f c ) tan( 2 f c )  tan(1 f c )  tan( 2 f c ))  j[tan( 1 f c )( 2 Y  tan( 2 f c )  tan( 3 f c ))  1]
2 j ( 2 Y tan(1 f c ) tan( 2 f c ) tan( 3 f c )  tan(1 f c ) tan( 2 f c )  tan( 2 f c ) tan( 3 f c )  tan(1 f c ) tan( 3 f c ))

2 (tan( 1 f c ) tan( 2 f c ) tan( 3 f c )  2 Y tan(1 f c ) tan( 2 f c )  tan(1 f c )  tan( 2 f c ))  j[tan( 1 f c )( 2 Y  tan( 2 f c )  tan( 3 f c ))  1]
(3.16)
The output impedance ̅̅̅̅̅
𝑜𝑑
𝑍𝑡𝑜𝑡 equals to 1 at the centre frequency 𝑓𝑐̅ for perfect matching at output
ports. Thus, the total impedance of the isolation circuit can be derived by (3.16). The impedance
̅̅̅̅̅̅
𝑍𝐼𝑆𝑂 can be calculated by substituting the design parameters of the transmission lines that
described in the previous section. Since a series RLC circuit is used in this design, the real part
and imaginary part of ̅̅̅̅̅̅
𝑍𝐼𝑆𝑂 should satisfy:



Re Z ISO 
R
2
 


Im Z ISO  2C L -

 1
2C C
 (3.17)
where ̅𝜔̅̅̅𝑐 is the normalized centre angle frequency. The relationship between design
parameters (such as the values for RLC) and perfect isolation performance among the circuit
ports can be found through analysing the network isolation using the method that was presented
in [4] [10] [11]. The summation admittances of both circuit branches are zero when the circuit
is entirely isolated.
Design Parameters and Simulation
Theoretical equations in Section 3.2.1 and Section 3.2.2 are derived to determine design
parameters. Firstly, the electrical length of the first and second transmission lines 𝜃1 and 𝜃2
should be determined based on the original centre frequency. The original centre frequency 𝑓0
is designed to be 1 GHz and all the terminal loads are 50 Ω. Firstly, 𝜃3 = 7°and 𝑍1 = 𝑍2 =
𝑍3 = 70.7 Ω are selected in the design. Then, to introduce an additional zero reflection
frequency 𝑓𝑍 at 1.8 GHz for a 20 dB bandwidth, the electrical length of the transmission
P a g e | 69
lines 𝜃1 = 53° , 𝜃2 = 37° (where 𝜃1 + 𝜃2 = 90° is assumed in this design) can be determined
using (3.11). Therefore, the capacitance 𝐶𝑃 can be calculated by substituting 𝜃1 = 53°, 𝜃3 =
7° and 𝑓𝑍 = 1.8 GHz into (3.10).
(a)
(b)
Fig. 3. 4 (a) AWR schematic circuit, (b) comparison of the simulated 20 dB bandwidth of
S11 between the proposed divider and previous works [1], [3] using AWR.
P a g e | 70
To work out the values of lumped elements on isolation circuit, 𝜃1 , 𝜃2 , 𝜃3 , 𝐶p and centre
(𝑓𝑜 + 𝑓𝑧 )⁄
frequency 𝑓𝑐 (where 𝑓𝑐 = 2 ) are substituted into (3.16) to obtain the total
impedance 𝑍𝐼𝑆𝑂 . Then according to (3.17), the real part of 𝑍𝐼𝑆𝑂 is half of the resistance while
the imaginary part of 𝑍𝐼𝑆𝑂 is the sum of the inductance and capacitance on the isolation circuit.
As a result, R, L and C can be obtained respectively. Fig. 3. 4(a) shows the schematic circuit
for the proposed PD. The component values of the schematic circuit can be obtained by using
(3.10),(3.11), (3.16) and (3.17). The return loss S11 can be calculated and shown as the blue
dash curve in Fig. 3. 4(b). The proposed design was simulated and optimized using AWR to
achieve a wider 20 dB bandwidth than a Trantanella-type PD or a conventional Wilkinson PD.
Then schematic circuit was realized using distributed elements of transmission lines. Sonnet
was used in this design. The PD was designed on a Rogers RT/5880 substrate with ɛr=2.2. The
thickness of the substrate is 0.79 mm, and the copper cladding is 35 μm thick. The layout is
shown in Fig. 3. 5. The front transmission lines are bent to fit in the capacitors. The widths of
all transmission lines are the same in this design. The lengths of the transmission lines were
finely adjusted for optimized results. All the calculated and optimized design parameters are
listed in Table 3.1.
Fig. 3. 5 Fabrication layout of proposed two-way PD.
P a g e | 71
Table 3. 1 COMPARISON OF CALCULATED AND OPTIMIZED PARAMETERS.
Items Cal. Opt. Items Cal Opt.
LTX1 35.9 mm 38 mm 𝑪𝑷 0.19 pF 0.2 pF
LTX2 19.8 mm 22 mm R 139Ω 130Ω
LTX3 6.3 mm 8 mm L 5 nH 3.9 nH
W 1.34 mm 1.35 mm C 1.6 pF 1 pF
3.3 Parametric Studies Regarding Bandwidth
The bandwidth of the proposed PD can be affected by choosing different parameters such as
𝜃1 (under the condition of 𝜃1 + 𝜃2 = 90°) and capacitance 𝐶𝑝 . Parametric studies regarding
bandwidth are conducted in this section. Although the bandwidth cannot be tuned in a linear
behaviour, the control of the bandwidth can be guided by sketching the bandwidth plot against
the variation of 𝜃1 and 𝐶𝑝 . Simulations are conducted using Matlab and AWR.
Bandwidth Regarding 𝜽𝟏
The bandwidth changes result from the additional reflection-zero. If the reflection-zero is
located far from the original reflection minimum (around 1 GHz) then the bandwidth could be
broadened significantly. Theoretically, the S11 response varies depending on different 𝜃1
( 𝜃1 + 𝜃2 = 90°). Fig. 3. 6 describes the additional reflection-zero 𝑓𝑧 changes as a function
of 𝜃1 and corresponding 𝐶𝑃 with all the other parameters fixed. All sets of design parameters
are calculated as described in Section 3.2.3, the curves in Fig. 3. 6 are plotted using Matlab.
The figure illustrated that a larger 𝜃1 can produce a reflection-zero at the higher frequency with
a decrease of 𝐶𝑃 . However, a trade-off should be made for bandwidth and overall in-band
performance. The ripple between 𝑓𝑜 and𝑓𝑐 is described as worst in-band S11. As it shown is
Table 3.2, the worst in-band S11 deviates with the increase of 𝜃1 .
P a g e | 72
Fig. 3. 6 Circuit parameters for realising different additional zero-reflection frequencies.
Table 3. 2 CALCULATED DESIGN PARAMETERS FOR DIFFERENT ADDITIONAL ZERO

REFLECTION FREQUENCIES.
Additional Zero Reflection 𝜽𝟏 Capacitance CP Worst S11

Frequency 𝒇𝒁 (GHz) (°) (pF) In-Band (dB)
1.2 34.2° 1.079 -50.4
1.4 41.6° 0.661 -30.11
1.6 48.0° 0.408 -26.2
1.8 53.0° 0.256 -19.6
2.0 56.8° 0.158 -17.02
2.2 59.6° 0.088 -15.11
2.4 61.5° 0.037 -13.56
2.5 62.1° 0.015 -12.86
A contour map is shown in Fig. 3. 7 to give a comprehensive understanding of 𝜃1 and the

corresponding bandwidth. The two concentrated areas indicate two S11 reflection-zeros that
locate at around 1 GHz and 1.8 GHz, respectively. Take -20 dB as the measure of bandwidth,
P a g e | 73
the achievable S11 bandwidth can be observed by drawing a horizontal line on the contours.
For example, a horizontal line can be drawn at 𝜃1 = 55° in Fig. 3. 7. This line has two
intersections with the 20-dB contour and a wide 20 dB bandwidth is achievable when 𝜃1 varies
from 54° to 58°. A much wider 15-dB bandwidth can be achieved when 𝜃1 varies from 44° to
65° with a broad distance between intersections on the 15-dB contour. Those intersections are
represented by 𝑓𝐻−20 𝑑𝐵 , 𝑓𝐿−20 𝑑𝐵 , 𝑓𝐻−15 𝑑𝐵 , and 𝑓𝐿−15 𝑑𝐵 in Fig. 3. 7, respectively. Hence the 20
dB S11 bandwidth can be read from the horizontal difference between 𝑓𝐻−20 𝑑𝐵 and 𝑓𝐿−20 𝑑𝐵 .
While the difference between 𝑓𝐻−15 𝑑𝐵 and 𝑓𝐿−15 𝑑𝐵 represents the 15-dB bandwidth. For 𝜃1 =
55°, the achievable 20 dB bandwidth is from 0.8 GHz to 1.9 GHz (bandwidth equals to2.37:1),
while the achievable 15 dB bandwidth is from 0.65 GHz to 2.1 GHz (bandwidth equals to
3.2:1).
Fig. 3. 7 The contour of S11 with frequency as a function of 𝜽𝟏 when 𝜽𝟐 = 𝛑⁄𝟐 − 𝜽𝟏 CP =

0.256 pF 𝐚𝐧𝐝 𝜽𝟑 = 𝟕°.
In practice, it is challenging to vary the value of 𝜃1 once the PD is fabricated. Thus, choosing
appropriate 𝜃1 at the first place is essential for the design. Fig. 3. 8(a)-(c) depict the bandwidth
contours regarding a changing 𝜃1 with 𝐶𝑝 = 0.2 pF, 0.4 pF and 0.8 pF, respectively. It can be
seen from Fig. 3. 8(a) when 𝐶𝑝 = 0.2 pF, although the two reflection minima are wide apart,
the 20 dB bandwidth is divided into two parts. In this case, a wide -20 dB bandwidth is not
P a g e | 74
(a)
(b)
(c)
Fig. 3. 8 The contour of S11 with frequency as a function of 𝜽𝟏 when 𝜽𝟐 = 𝛑⁄𝟐 − 𝜽𝟏 , 𝜽𝟑 =

𝟕° and (a) 𝑪𝒑 =0.2 pF, (b) 𝑪𝒑 =0.4 pF and (c) 𝑪𝒑 =0.8 pF.
P a g e | 75
achievable no matter what value of 𝜃1 is chosen, but the 15 dB bandwidth could be comparably
large. So compromise is needed. Fig. 3. 8(b) shows that the two dark areas are falling for a
smaller 𝜃1 which means a smaller 𝜃1 is sufficient to produce two reflection minima, hence
realising the widest bandwidth. Fig. 3. 8(c) illustrates that by further increasing 𝐶𝑝 , a better in-
band performance but a narrower bandwidth can be achieved with even smaller 𝜃1 . To
summarize Fig. 3. 8, generally speaking, a smaller 𝐶𝑝 can provide a larger 20 dB bandwidth
when 𝜃1 is around 45° to 50°. While using a higher 𝐶𝑝 can obtain a narrower bandwidth but a
better in-band performance with 𝜃1 around 35°.
Bandwidth Regarding Capacitance 𝑪𝒑
This section studies the effects of different 𝐶𝑝 with a fixed 𝜃1 , 𝜃2 and 𝜃3 . For a conventional
quarter-wavelength Wilkinson PD, the reflection zero locates at the centre frequency of 1 GHz.
This reflection-zero becomes a reflection minimum in the proposed design due to the added
capacitors. As mentioned above, the location of the additional reflection-zero determines the
relationship between bandwidth and capacitance 𝐶𝑝 . The larger the distance between the
additional reflection zero and the reflection minimum (slightly larger than 1 GHz), the wider
the bandwidth could be. However, that is not always the case because the in-band ripple will
get higher with the increase of 𝑓𝑧 . The bandwidth is limited by the ‘worst’ in-band S11 which is
the peak value of the in-band ripple. As is discussed in the previous section, each 𝜃1
corresponds to a single value of 𝐶𝑃 that can produce an additional reflection-zero 𝑓𝑧 but the
bandwidth can still be changed by tuning 𝐶𝑃 .
P a g e | 76
Fig. 3. 9 Changing CP produces varying 𝐒𝟏𝟏 𝐚𝐧𝐝 𝐒𝟐𝟏 with 𝜽𝟏 = 𝟓𝟑°, 𝜽𝟐 𝟑𝟕° 𝐚𝐧𝐝 𝜽𝟑 = 𝟕°.
Some design examples are shown in Fig. 3. 9. The original centre frequency is assumed to be 1
GHz, and the additional transmission minimum varies from 1.2 GHz to 2.5 GHz by
changing 𝐶𝑝 under the condition of 𝜃1 = 53°, 𝜃2 = 37° and 𝜃3 = 7°. The reflection coefficient
at the original centre frequency 1GHz is no longer zero (-∞ in dB) by adding a pair of
capacitors 𝐶𝑃 , but the return loss is still very low (< -20 dB). The reflection minimum frequency
moves slightly from the original centre frequency (1GHz), but it is still very close to 1 GHz
(1 < 𝑓 ̅< 1.12) according to (3.3), (3.4). It can be seen that an additional reflection-zero is
generated by adding a 𝐶𝑃 = 0.256 pF, and the in-band ripple is exactly on 20 dB. When 𝐶𝑃 is
increasing, the ripple will descend but the bandwidth will shrink. A wider bandwidth can be
achieved by reducing 𝐶𝑃 but the in-band performance will deteriorate simultaneously. In other
word, for a given layout, increasing 𝐶𝑃 will obtain a better in-band performance with either a
narrower bandwidth or a wider bandwidth while a compromised in-band performance can be
achieved by decreasing 𝐶𝑃 .
P a g e | 77
Fig. 3. 10 The contours of S11 as a function of frequencies and CP when 𝜽𝟏 = 𝟓𝟑° and
𝜽𝟑 = 𝟕°.
To emphasize the bandwidth and in-band performance changing in terms of different 𝐶𝑃 , Fig.
3. 10 shows the contours of S11 as a function of the capacitance 𝐶𝑃 with 𝜃1 = 53° and 𝜃3 = 7°.
The two dark areas represent the two low-reflection frequency zones at around 1.1 GHz and 1.8
GHz in this case. The -30 dB, -20 dB, and -15 dB contours are shown in the figure. When 𝜃1 =
53° and 𝜃3 = 7°, the calculated 𝐶𝑃 equals to 0.255 pF, as shown in Fig. 3. 10, S11 is slightly
worse than 20 dB at around 1.4 GHz. If 𝐶𝑃 is chosen to be 0.4 pF, S11 is better than -20 dB from
𝑓𝐿−20 𝑑𝐵 = 0.8 GHz to 𝑓𝐻−20 𝑑𝐵 = 1.6 GHz. Slightly decreasing 𝐶𝑃 , the distance between
𝑓𝐿−20 dB and 𝑓𝐻−20 dB can be further extended. On the contrary, the bandwidth will shrink when
𝐶𝑃 increases. Therefore, the bandwidth of the proposed divider can be controlled in a wide range
by alternating the capacitance 𝐶𝑃 . It can be concluded that when the capacitance of 𝐶𝑃 is small
(but not smaller than 0.255 pF), the bandwidth is large while if 𝐶𝑃 is increasing towards 0.8
pF, then the passband will attenuate. However, if -15 dB S11 is sufficient for the design, the
tunable range of 𝐶𝑃 will be large. Also, if 𝐶𝑃 is smaller than 0.2 pF, then the performance will
become a dual-band response. It can be concluded that for a fixed layout, 𝐶𝑃 has a value range
to maintain an effective bandwidth, and the bandwidth can vary from zero to maximum when
P a g e | 78
𝐶𝑃 is in this range. If a tunable bandwidth is preferred, a varactor can be employed to the circuit
to have reconfigurable responses.
To sum up, if both 𝐶𝑃 and 𝜃1 are considered in the design, the theoretical maximum 20 dB
bandwidth has the trend as shown in Fig. 3. 11. The maximum bandwidth in this figure is 2.32:1
with 𝐶𝑝 = 0.25pF and 𝜃1 = 55°. In general, a small 𝐶𝑝 and a large 𝜃1 can contribute to a wide
bandwidth, but an overlarge 𝜃1 will result in a worse in-band performance which limits the
extension of the bandwidth.
Fig. 3. 11 The maximum 20 dB bandwidth regarding 𝜽𝟏 and corresponding 𝑪𝒑 .
3.4 General Cases
In the previous sections of this chapter, impedances 𝑍1 , 𝑍2 and 𝑍3 are assumed to be 70.7 Ω
and 𝜃1 + 𝜃2 = 90° is defined for the ease of calculation. However, it sacrifices the freedom of
the design. Generally, for the proposed PD, the impedances and electrical lengths of
𝑍1 , 𝑍2 and 𝑍3 could be different from each other instead of being 70.7Ω. Moreover, the sum of
𝜃1 and 𝜃2 is not necessarily to be 90°. The calculation process for the capacitance 𝐶𝑝 is the
same with (3.3), but the impedances 𝑍1 , 𝑍2 and 𝑍3 are variables instead of being √2𝑍0 . Thus,
the relationship between these parameters can be derived as:
P a g e | 79
sin(θ2 𝑓̅) − cos(θ1 𝑓̅)YZ1 cos(θ2 𝑓)̅ − 2 sin(θ2 𝑓)YZ

̅ 1 cos(θ2 𝑓̅)YZ1 − sin(θ2 𝑓̅)
+ +
tan(θ1 𝑓)̅
2
tan(θ1 𝑓)̅ sin(θ1 𝑓)̅
2
=
A+B
cos(θ2 𝑓)̅ Z1 − sin(θ2 𝑓 ̅) YZ12 − 2 sin(θ2 𝑓 ̅) Z2 2 sin(θ2 𝑓 ̅)YZ1 Z2 − 2 cos(θ2 𝑓)Z

̅ 2 − sin(θ2 𝑓)Z
̅ 1 2 sin(θ2 𝑓 ̅)Z2 + sin(θ2 𝑓 ̅)YZ12 − cos(θ2 𝑓 ̅)Z1
+ +
tan(θ1 𝑓 ̅)2 tan(θ1 𝑓 ̅) sin(θ1 𝑓 ̅)2
̅ ̅
(B − A − cos(θ2 𝑓 )) Z2 − sin(θ2 𝑓 ) Z1
(3.18)
̅)−cos(θ1 𝑓
sin(θ2 𝑓 ̅)YZ1 cos(θ2 𝑓 ̅)YZ1 cos(θ2 f̅)YZ1 −sin(θ2 𝑓
̅)−2 sin(θ2 𝑓 ̅)
̅ 2 + ̅ + ̅ 2
tan(θ1 𝑓) tan(θ1 𝑓) sin(θ1 𝑓)
𝑌= (3.19)
Z1 (A+B)
A = cos(θ2 𝑓)̅ − 2 sin(θ2 𝑓)YZ

̅ 1
4 𝑠𝑖𝑛(𝜃2 𝑓̅)(1−𝑐𝑜𝑠(𝜃2 𝑓̅))

B=
𝑠𝑖𝑛(𝜃1 𝑓̅)
tan(θ3 𝑓̅ )
α=
Z3
To find suitable 𝐶𝑝 , all the parameters have to satisfy (3.11) at a selected reflection-zero
frequency 𝑓𝑧 . Then, Y can be calculated using (3.12), hence 𝐶𝑝 = Y⁄(2𝜋𝑓̅𝑧 ) can be found. It
should be noted that although the length of the transmission line 𝑍3 is much shorter compared
to 𝑍1 and 𝑍2 , the changes of α can also affect the input impedance of the circuitry. So the
̅) and 𝑍 have a ratio of α, if θ changes, then
changes of α cannot be ignored. Since tan(θ3 𝑓 3 3
the value of 𝑍3 needs be changed to maintain the same α. By doing so, the divider can achieve
an identical performance. The calculation for the parameters in general cases is much
complicated than the calculation in the previous section. It is preferred to substitute predefined
parameters into the equations for simplification.
3.5 Experimental Results
Fabrication and Measurement
A two-way PD is fabricated and measured as shown in Fig. 3. 12. LPKF ProtoMat S62 is used
for circuit fabrication. The circuit layout was exported from Sonnet and imported to the easy-
P a g e | 80
to-use LPKF CircuitCAM software for working process. The layout is printed on top of a
Rogers RT/5880 0.031” substrate while the bottom is covered by copper as the ground of the
circuit. The fabricated circuit has a compact size of 29 mm × 32 mm while the two output ports
can be placed widely apart. A pair of surface amount capacitors and R, L, C on the isolation
circuit are soldered on to the circuit. Three SMA connectors are soldered to the ports for
measurement.
The circuit is measured by using a VNA (vector network analyser) Agilent FieldFox N9917A.
The frequency sweeping range of the VNA is from 30 KHz to 18 GHz which is suitable for the
proposed design. Firstly, the NA (network analyzer) mode is chosen to operate the equipment.
Then S-parameter measurement was carried out. Before recording data, the measurement
system (the VNA with two SMA cables) is calibrated by using a SOLT (Short-Open-Load-
Thru) calibration kit. S11, S21 and S22 can be measured by connecting cables to Port 1 and Port
2. S31 and S33 are measured by connecting cables to Port 1 and Port 3. The isolation S23 is
measured between Port 2 and Port 3. A 50 Ω male load is always connected to the spare port
that is not measured. The measurement was repeated several times, and the precision is around
± 0.1 dB. The measured magnitude of S-parameters versus sweeping frequencies are saved in
corresponding files which can be imported to Matlab or Origins to plot measured results.
Fig. 3. 12 Photographs of the measurement set up and the fabricated two-way divider.
P a g e | 81
Two-way PD
The measured results of the proposed PD with 𝐶𝑃 =0.2 pF are shown in Fig. 3. 13. It can be
seen that a reflection-zero is generated as expected, and the operational bandwidth of the
divider is significantly increased. As analyzed in Section 3.3.2, 𝑓𝑍 can be controlled by 𝐶𝑃 .
Two sets of S11 curves that correspond to 𝐶𝑃 = 0.4 pF and 𝐶𝑃 = 0.8 pF are also shown in Fig.
3. 13(a). When 𝐶𝑃 = 0.8 pF, there is only one reflection minimum that locates at around 1 GHz
which results in a narrow bandwidth. Decrease the capacitance of 𝐶𝑃 to 0.4 pF, an additional
reflection minimum appears in the upper passband at around 1.6 GHz, and the measured 20 dB
bandwidth is 2.1:1. Then continually reducing the capacitance of 𝐶𝑃 to 0.2 pF, the additional
zero-reflection frequency moves to around 1.9 GHz. The 20 dB bandwidth cannot be realised
because the in-band ripple is worse than 20 dB. The measurement shows a 15 dB bandwidth
of 3.5:1 with 𝐶𝑃 equal to 0.2 pF. This is much greater than that of traditional Wilkinson PDs or
typical Trantanella ones. The measured reflection and isolation performances are also very
good (<-15 dB) in the desired operational band. The measured insertion loss S21 is flat at around
3 dB over the whole operational band as Fig. 3. 13(b) shows. The measured results are in very
good agreement with the simulations shown in Fig. 3. 14(a) (b). Table 3.3 compares the
measured performances of the two-way with other similar works. For two-way dividers, the
effective bandwidth of the proposed one is much wider than all other works.
P a g e | 82
(a)
(b)
Fig. 3. 13 Measured results of (a) S11 & S21, (b) S22 & S23 for proposed two-way PD.
P a g e | 83
(a)
(b)
Fig. 3. 14 Simulated results of (a) S11 & S21, (b) S22 & S23 for proposed two-way PD.
P a g e | 84
Table 3. 3 PERFORMANCE COMPARISON OF TWO-WAY WILKINSON-TYPE POWER

DIVIDERS WITH OTHER WORKS.
Reference Circuit Size Physical Isolation Bandwidth (15 dB)
[1] 0.25λg1 *0.02λg NO 2:1
[2] 0.2λg *0.05λg YES NA
[3] 0.25λg*0.06λg YES 1.45:1
[5] 0.34λg *0.2λg YES 1.85:1
This work 0.12λg *0.1λg YES 2.73:1

1: λg is the guided wavelength at the centre frequency of the bandwidth.
Eight-way PD
This method can also be applied to design multi-way PDs. The challenges are the complexity
of design parameter calculations, and how to avoid deterioration of in-band performance by
cascading more stages with such a wide bandwidth. An eight-way PD is designed and
fabricated on the same substrate for further demonstration as shown in Fig. 3. 15.
Fig. 3. 15 The fabricated circuit of eight-way PD.
P a g e | 85
(a)
(b)
Fig. 3. 16 (a) Measured response and (b) simulated results of S11, S21, S22 and S23 for the
eight-way PD.
The eight-way PD has three stages so that analytical solutions for design parameters are more
difficult to work out. To simplify the calculation, the design parameters in every single section
of the eight-way divider are almost identical (except for the lumped elements) to the two-way
divider, which means no further changes on transmission lines are needed for multi-way PDs.
P a g e | 86
(a) (b)
(c) (d)
(e) (f)
Fig. 3. 17 (a) S11, S41/S51, and (b) S44, S45 and S55 for ports 4&5, (c) S11, S61/S71, and (d) S66,
S67 and S77 for ports 6&7, (e) S11, S81/S91, and (f) S88, S89 and S99 for ports 8&9.
P a g e | 87
The values of 𝐶𝑃 and the lumped elements for the isolation circuits in different stages are
optimized by simulation. The measured and simulated results for the eight-way PD between
Port 1 to Port 2 & 3 can be seen in Fig. 3. 16(a) and (b), respectively. The simulated S11 is
below 17 dB from 0.5 GHz to 2 GHz while the return losses, S22/S33 are all below 18 dB over
0.6 GHz to 2 GHz. The isolation S23 is better than 16 dB over 0.75 GHz to 1.9 GHz. The
measured return loss S11 and the output reflection S22 are better than 15 dB while the measured
isolation S23 remains better than 13 dB. Compared with the simulation, the frequency shift and
deviation of S-parameters are mainly caused by the fabrication error and the inaccuracy of the
lumped elements. The rest ports (ports 4&5, ports 6&7 and ports 8&9) are also tested, and the
measured results are plotted in Fig. 3. 17(a)-(c). Ideally the results should be identical, however,
due to the fabrication error and cable rotations, the measured results are slightly different.
Overall, the bandwidth of the eight-way PD is wider than previouse similar works [12]-[15].
The measured results show very good agreement with the simulated ones.
3.6 Summary
Compact and wideband PDs are more and more desirable in modern microwave
communication systems. The Trantanella-type PDs with the isolation circuit moved from the
output ports to the middle of the transmission line can provide a compact size and additional
physical isolation. A new design method has been proposed in this chapter that a pair of
capacitors are added in parallel with the front-section transmission lines to broaden the
operational bandwidth. The added capacitors have significantly increased the bandwidth of the
PD. The design guidance regarding choosing the lengths and characteristic impedances of
transmission lines, and the capacitance 𝐶𝑝 are discussed in detail. The bandwidth can be
enlarged by properly choosing design parameters, but the increment is limited by the
deteriorated in-band performance. Experiments have been conducted to verify the design. For
the proposed two-way PD, the measured 15-dB-return-loss bandwidth of the two-way divider
is 3:1 while the effective bandwidth regarding all the S-parameters (S11, S21, S22, S33 and S23) is
2.73:1. For the eight-way PD, the measured 13-dB bandwidth is 2.67:1 considering return
losses and isolation while the return loss bandwidth can achieve 4.5:1. The simulated and
measured results confirm that this proposed method can provide a broadband response
compared to other works. Also, changing the added capacitors will result in a broad tunable
P a g e | 88
bandwidth (from 0 to 2.73: 1) of the dividers. Apart from the extensive and tunable bandwidth,
the proposed divider also has a compact size and can provide physical separation as well as
electrical isolation among output ports. Moreover, the measurement on the eight-way PD has
proved that this proposed method can be easily extended to multi-way divider designs.
References
118, 19060.
[2] C-.J-.Trantanella, “A novel power divider with enhanced physical and electrical port isolation,”
IEEE MTT-S Int., pp. 129-132, 2010.
[3] X. Wang, I. Sakagami, A. Mase, and M. Ichimura, “Trantanella Wilkinson power divider with
additional transmission lines for simple layout.,” IET Microwaves, Antennas & Propagation, vol.
8, no. 9, pp. 666-672, 2014.
[4] S.-W. Lee, C.-S. Kim, K. S. Choi, J.-S. Park, and D. Ahn, “A general design formula of multi-
section power divider based on singly terminated filter design theory,” in IEEE Microw. Symp.
Dig., Jun. 2001, pp. 1297–1300.
[5] K. Song and Q. Xue, “Novel ultra-wideband (UWB) multilayer slotline power divider with
bandpass response,” IEEE Microw. Wireless Compon. Lett., vol. 20, no. 1, pp. 13–15, Jan. 2010.
[6] M. A. Maktoomi, M. S. Hashmi, and F. M. Ghannouchi, “Theory and design of a novel wideband
DC isolated Wilkinson power divider,” IEEE Microw. Wireless Compon. Lett., vol. 26, no. 8, pp.
586–588, Aug. 2016.
[7] W.-J. Feng, Y. Zhang and W.-Q. Che, “Tunable Dual-Band Filter and Diplexer Based on Folded
Open Loop Ring Resonators,” IEEE Trans. Circuits Syst. II Exp. Briefs, vol. 64, no. 9, pp. 1047-
1051, 2017.
[8] T.Zhang and W.-Q. Che, “A compact Tunable Power Divider With Wide Tuning Frequency Range
and Good Reconfigurable Resources,” IEEE Trans. Circuit Syst.II, Exp. Briefs, vol. 63, no. 11, pp.
1054-1058, 2016.
[9] W.-D. Lin, R.-C. Liu, B.-L. Chen, P.-H. Deng and S.-F. Chao, “A Wilkinson power divider with
transmission zero in desired stopband using embedded parallel resonator,” in Microwave
Conference (APMC), 2014 Asia-Pacific, Sendai, Japan , 2015.
[10] S.-Y. Hu, K. Song, and Y. Fan, “Planar ultra-wideband eight way power divider with improved
isolation bandwidth,” in Microwave Confrence (APMC), Asia-Pacific. vol 2. IEEE, 2015.
P a g e | 89
[11] Mohammad A. M, Mohammad S. H and Fadhel M. G, “Theory and Design of a Novel Wideband
DC Isolated,” Microwave and Wireless, vol. 26, no. 8, pp. 586-588, 2016.
[12] M. Elsbury et al., “Integrated broadband lumped-element symmetrical hybrid N-way power
dividers,” in IEEE IMS Dig., Jun. 2009.
[13] C. Zhu, W. Kang and W. Wu, “Quasi-lumped dual-mode reconfigurable eight-way FPD,”
Electronic Lett. Vol: 53, no. 25, pp1657-1659, 2017.
[14] R. Kazemi and A. E. Fathy, ‘‘Design of a wideband eight-way single ridge substrate integrated
waveguide power divider,’’ IET Microw. Antenna Propag., vol. 9, no. 7, pp. 648–656, Jul. 2015.
[15] T. Yu, “Design of Length-Saving Multiway Wilkinson Power Dividers,” IEEE Access, vol:6,
pp:14093-14105. March, 2018.
P a g e | 90
Chapter 4: A Horst-type PD with Wide Frequency Tuning Range Using Varactors
Chapter 4 A Horst-type PD with Wide Frequency

Tuning Range Using Varactors
This chapter presents a novel PD with a wide frequency tuning range. In the previous work that
presented in Chapter 3, a pair of capacitors were connected in parallel with the front transmission
lines of a Trantanella-type PD introducing an additional reflection minimum together with the
original reflection zero to broaden the bandwidth. In this design, the latter transmission lines are
removed. The added capacitors generate a reflection minimum which can be easily controlled by
varactors. Thus, the frequency band of the PD can be flexibly tuned by altering the varactors.
Design parameters are carefully chosen to eliminate the effect of the original reflection zero. By
doing so, the PD will have a tunable centre frequency instead of a tunable bandwidth. Theoretical
formulas for the characteristic impedances and electric lengths of the transmission lines of the
PD are derived and analyzed. A PD has been designed and fabricated to demonstrate the validity
of the proposed design. The measured results indicate that the PD can achieve a frequency tuning
range of 0.9-4.2 GHz (𝒇𝑯 ⁄𝒇𝑳 = 4.67:1), with in-band input and output return losses both better
than 22 dB, and an insertion loss S21 of 3.2-4 dB. The measured in-band isolation is better than
15 dB. The PD has a simple layout and a compact size of 0.2𝝀𝒈 ×0.16𝝀𝒈 which demonstrates
the excellent potential of the proposed PD for modern communication systems.
4.1 Introduction
Modern communication systems usually need to support multi-standard communications,

which require the components with multiple operational bands. PDs are essential components
in communication systems for power splitting and combining. The features of a PD regarding
the operational frequency band, input/output return loss, and isolation between the output ports
will significantly affect the overall performance of the communication system. Although many
designs [1], [2], [3] of multi-band PDs have been reported, their circuit size is usually bulky.
Modern wireless communication prefers tunable RF components instead of multiplexing into
separated frequency bands due to the size and cost reduction. As the demand for spectrally
cognitive microwave system is rising, more and more researches are focusing on frequency
P a g e | 91
agile devices, such as reconfigurable couplers [4], resonators and filters [5]-[9]. With the
rapidly increasing demands on multi-band feature of the PDs, there is growing attention paid
on PDs with compact size and wide tunable frequency ranges. Filtering tunable PDs [10]-[12]
have been investigated over the past years. In these designs, a filter and a PD are cascaded to
achieve filtering PD response. The filter replaces the quarter-wavelength transmission lines in
Wilkinson PDs. The main contribution of such designs is to provide a filtering response to the
PDs. However, there are drawbacks. The insertion loss is relatively high (1.8-2.4 dB) due to
the limited unloaded quality factor of microstrip resonators. Moreover, the port matching and
isolation conditions are usually not satisfactory. To maintain better performance in the entire
tunable frequency band, [13] [14] introduced coupled-line tunable Wilkinson PDs which have
excellent input and output matching (|𝑆11 |, |𝑆22 | < −17 dB) and isolation (|𝑆23 | < −25 dB).
The aforementioned PDs are good to be used in many filtering PD applications. However, the
limitation of the 20 dB tuning range (normally less than 2:1) is still a challenge. Modern
transceivers like software defined radios require a wide range of radio spectrum. The designs
in [15] and [16] have a tunable frequency range of 2.9:1 by employing short-electrical
transmission lines along with a shunt varactor to the Wilkinson-type PD, but no analytical
solutions to describe the relationship of capacitance and the corresponding frequency were
provided.
A novel PD with a wide frequency tuning range and a compact footprint is presented in this
section. Similar to [17], a pair of varactors are introduced in parallel with the front transmission
lines. The main difference is that the design in [17] has an extra pair of transmission lines that
connect the outputs with the isolation networks so that the varactors can tune the frequency
tuning range of the divider. The proposed divider in this work utilises a Horst Divider structure
[18], [19] to simplify the calculation. In this design, the extra transmission lines were removed
to eliminate the original reflection-zero in [17] and maintain the additional reflection minimum
which is introduced by the varactors. As a result, the proposed design can achieve a very wide
frequency tuning range. The reflection minimum is very sensitive to the capacitance change.
The centre frequency of the operational band can be defined as the frequency of the
corresponding reflection minimum. The centre frequency moves along a locus. In other words,
the centre frequency tuning range can be controlled flexibly by the varactors. An example in
this chapter demonstrated an available frequency tuning range of 0.9-4.2 GHz by tuning the
P a g e | 92
varactors. Excellent output return loss and isolation response are retained over the whole tuning
range.
Moreover, good physical isolation between two output ports can be realised by using a Horst
structure which is a pair of short-length-transmission lines connected between the front
transmission lines and the isolation circuit. Excellent physical isolation not only separates two
output ports but also suppresses the undesirable coupling between them. The design equations
and analysis of the proposed structure are presented in Section 4.2. Then the parametric studies
regarding tuning range are discussed in Section 4.3. In Section 4.4, a prototype of the proposed
PD is fabricated and measured to validate the proposed theory. In the end, conclusions will be
drawn in Section 4.5.
4.2 Analytical Design Equation for the Proposed Horst-type PD
Fig. 4. 1 Structure of the proposed Horst-type PD.
Fig. 4. 1 presents the equivalent circuit of the proposed two-way reconfigurable PD. A pair of
varactors are added in parallel to the main transmission lines to realise tunability, and a single
varactor is inserted in the isolation circuit to compensate the output return losses and isolation.
As Fig. 4. 1 shows, 𝑍0 represents the port impedance. 𝑍1 and 𝑍2 are the characteristic
impedances of the front transmission lines and the transmission lines in the isolation circuit,
P a g e | 93
respectively. 𝐶1 represents the capacitance of the added varactors on the front transmission
lines, 𝐶2 is the capacitance of the varactor on the isolation circuit connected in series with a
resistor. The electrical lengths of the transmission lines are defined as 𝜃1 and 𝜃2 at frequency f
in the same way as 𝜃0 at the centre frequency𝑓0 . The relationship among the characteristic
impedances, electrical lengths of the transmission lines and the varactors can be derived by an
even-mode analysis method. To design the isolation circuit, the odd-mode analysis method can
be applied to calculate the lumped resistor R and varactor 𝐶2 .
Even-mode Analysis
Fig. 4. 2 The equivalent circuit of the proposed structure for even-mode analysis.
The even-mode equivalent circuit of the proposed PD is depicted in Fig. 4. 2. The normalised
ABCD matrix of the structure is used to calculate the total input impedance of the circuit. The
normalised ABCD matrix for the network of the front transmission line with parallel-loaded
capacitors can be obtained referring to (3.4) in Chapter 3.
𝑗
𝑗𝑌̅−̅̅̅̅ ̅) 1
𝑍1 ∙tan(𝜃1 𝑓
𝑗 𝑗
| 𝑗𝑌̅−̅̅̅̅ 𝑗𝑌̅−̅̅̅̅ |
̅ 𝐵̅ | ̅) ̅)
|𝐴
𝑍1 ∙sin(𝜃1 𝑓 𝑍1 ∙sin(𝜃1 𝑓
= 2 2
𝐶̅ ̅ 𝐶 +Tx1 |(𝑗𝑌̅−
𝐷 𝑝 ̅̅̅̅
𝑍 ∙tan(𝜃
𝑗
̅
𝑓 )
) −(𝑗𝑌̅−̅̅̅̅
𝑍
𝑗
∙sin(𝜃 ̅ )) 𝑗𝑌̅−̅̅̅̅
𝑗
̅) |
1 1 1 1𝑓 𝑍1 ∙tan(𝜃1 𝑓
𝑗 𝑗
𝑗𝑌̅+̅̅̅̅ ̅ 𝑗𝑌̅−̅̅̅̅ ̅)
𝑍1 ∙sin(𝜃1 𝑓) 𝑍1 ∙sin(𝜃1 𝑓
(4.1)
Where Y  2 f  C 1
The ABCD matrix for the other transmission line can be expressed as
P a g e | 94
̅ 𝐵̅ 1 0
| 𝐴 | = |𝑗 ̅
1| (4.2)
tan(𝜃 3 𝑓 )
𝐶̅ 𝐷̅ Tx2 ̅̅̅
𝑍3
 1 
 Y 
Z1 tan(1 f ) 1
 
 Y
1  1 
 j Y -   0
Z1 sin(1 f ) 1
A B   Z1 sin(1 f )     (4.3)
  2 2 
  j tan( 2 f ) 1
C D   Y - 1  

1 
Y
1   
  Z1 sin(1 f )    
- Y Z2
Z1 tan(1 f )  Z1 tan(1 f ) 
 
  1  Y-
1 
 j Y -  Z1 sin(1 f ) 
  Z1 sin(1 f )  
where the frequency f is normalised to 1 GHz. The terminal impedance 𝑍0 is normalised to 1,

impedances 𝑍1 and 𝑍2 are normalised to 𝑍0 , respectively. The normalised ̅̅̅
𝐶1 equals to 𝐶1 ∙ 𝑍0 .
The normalised input impedance of the circuit can be calculated by using the ABCD matrix as:
A 1  B

even
Z in (4.4)
C 1  D
And the magnitude of the return loss can be expressed as:
A  B /1  C 1  2  D
c2  d 2
| S11 |  (4.5)
A  B /1  C 1  2  D a2  b2
where
a 
 1-2PY  Z 1
2
 
 2 sin(1 f )  Z1 4Y  2 P cos(1 f )  4Y  

 3Y  P  Z sin( f )  3cos( f )Z
2 2
b 
 1 1 1 1

c 

 2PY  1 Z  2 sin( f )  Z  4Y   4Y  2P  cos( f ) 
1
2
1 1 1

 P  Y  Z sin( f )  Z cos( f )
2
 d
 1 1 1 1
tan(2 f )
P
Z2
P a g e | 95
The |S11| in (4.5) equals to zero if the input terminal is perfectly matched at 𝑓̅𝑍 . Hence, the
𝑍2 and the zero-reflection frequency 𝑓̅𝑍 can be expressed
̅̅̅1 ,̅̅̅̅
relationship between impedance 𝑍
by:
 2 
 2 P  8P  2  sin(1 f Z )
Z1   
 
(4.6)

2
cos(1 f Z )  1  2 P  1
(4.6) gives the equation for determining ̅̅̅

𝑍1 and̅̅̅̅
𝑍2 , and values that satisfy the equation can
produce a reflection-zero at 𝑓̅𝑍 . Let 𝜃1 and 𝜃2 be independent variables that varying from
0° to 90° . Since 𝜃1 is the electrical length of the main transmission line, it can be assigned to
determine the proper tuning band. While 𝜃2 should be chosen from small values such as
0° ≤ 𝜃2 ≤ 20° to simplify the layout for practical applications. Moreover, it is necessary to
bend the front transmission lines for placing the varactors in parallel with them, and to achieve
a compact size of the topology. Hence, the ratio of length and width for the front transmission
lines need to be sufficiently large for bending. Specifically, a larger 𝜃1 indicates longer
̅̅̅1 implies narrower width of the transmission lines. Therefore,
transmission lines while lager 𝑍
𝜃1 and ̅̅̅
𝑍1 should be chosen appropriately for fabrication purposes.
Odd-mode Analysis
Fig. 4. 3 The equivalent circuit of the proposed structure for the odd-mode analysis.
P a g e | 96
To determine the parameters in the isolation circuit, odd-mode analysis is adopted. The odd-
mode equivalent circuit of the PD is shown in Fig. 4. 3. The normalized input impedance for
the odd-mode equivalent circuit is denoted by̅̅̅̅̅̅̅
odd
𝑍in , and the total impedance can be seen as
̅̅̅̅̅̅
𝑍1odd and ̅̅̅̅̅̅
𝑍2odd is in parallel.
𝑗𝑍 1 tan(𝜃1 ̅̅̅̅
̅̅̅̅ 𝑓𝑧 )
̅̅̅̅̅̅
𝑍1odd =
̅
𝑗𝑌
(4.7)
̅̅̅1̅ tan(𝜃1 𝑓̅𝑧 )+ 1̅
𝑗𝑍 𝑗𝑌
̅̅̅̅̅̅̅̅
𝑍ISO
+𝑗𝑍 ̅̅̅2̅ tan(𝜃2 𝑓̅𝑧 )
̅̅̅̅̅̅
𝑍2odd = ̅̅̅
𝑍2 2
(4.8)
̅̅̅̅̅̅̅̅
̅𝑍̅̅2̅+𝑗 𝑍𝐼SO tan(𝜃2 𝑓̅𝑧 )
2
𝑍ISO = 𝑅̅ ⁄2 + 1⁄(2𝜋𝑓̅𝑧 ̅̅̅

where ̅̅̅̅̅̅ 𝐶1 ). The input impedance of Port 2 can be derived.
̅̅̅̅̅̅̅ ̅̅̅̅̅̅̅
̅̅̅̅̅̅
odd 𝑍1odd ∙𝑍2odd
𝑍in = ̅̅̅̅̅̅̅
odd ̅̅̅̅̅̅̅
odd
(4.9)
𝑍1 +𝑍2
To guarantee perfect impedance matching at outputs, the following conditions should be

satisfied.
 Re  Z odd   1 (4.10)
in

Im  Z in   0
odd
By substituting design parameters to (4.10), the total impedance of the isolation circuit can be
calculated as:
2Z 2 [( j  C1 )Z1 Z 2 tan(1 f Z )2  Z1 tan(1 f Z )  Z 2 tan(2 f Z )]

ZISO  (4.11)
Z1 Z 2 tan(1 f Z )  j (Z1 tan(1 f Z )2  C1 Z1 Z 2 tan(1 f Z )  Z 2 )
Thus, the capacitance ̅̅̅

𝐶2 and resistance 𝑅̅ in the isolation circuit can be obtained by solving:

 Re Z ISO  R 2
 


 
Im Z ISO  1 2 C2 (4.12)
P a g e | 97
Since̅̅̅̅
𝑍1 ,̅̅̅̅
𝑍2 , 𝜃1 and 𝜃2 have been chosen from the even-mode analysis, hence a proper range
of ̅̅̅̅
𝐶2 can be determined by substituting different values of capacitance ̅̅̅
𝐶1 to (4.11).
4.3 Design Method
To extend the matching bandwidth, the value of ̅̅̅

𝑍1 could be changed from the calculated result
of (4.6). For a given frequency̅̅̅
𝑓𝑧 , values of ̅̅̅ ̅̅̅2 that satisfy equation (4.6) can generate
𝑍1 and 𝑍
one reflection-zero at 𝑓𝑧̅ . However, if ̅̅̅
𝑍1 changes, |S11| will not be equal to zero at the
frequencies near̅̅̅
𝑓𝑧 , but it can be kept very low so that the frequency tuning range can be
extended. The following sections will discuss the effect on the tuning range in terms of the
impedance̅̅̅̅ ̅̅̅2 and capacitance ̅̅̅
𝑍1 , 𝑍 𝐶1 . Also, a design guidance is provided at the end of this
section.
Determination of Impedance ̅̅̅

𝐙𝟏 and ̅̅̅
𝐙𝟐
For the proposed reconfigurable PD, continuously changing the capacitance ̅̅̅
𝐶1 will result in
consecutive S11 pass bands. Every single response band has a frequency minimum, ̅̅̅
𝑓𝑚
(normally the centre frequency of the tunable band). Thus, continuously tuning ̅̅̅
𝐶1 can produce
consecutive frequency minima which form an envelope curve of |S11|. The frequency tuning
range can be determined referring to the envelope curve as it records the movement of the |S11|
trough over the whole operating band. To improve the impedance matching range, the input
impedance should approximately equal to the load impedance over a wide frequency range.
This section discusses how to enhance the adaptive frequency range by choosing appropriate
values of ̅̅̅̅
𝑍1 and ̅̅̅̅
𝑍2 .
P a g e | 98
Fig. 4. 4 Theoretical reflection-zero frequency curve with changing Z1 and Z2.
Taking 𝜃1 = 30𝑜 , 𝜃2 = 18𝑜 and ̅̅̅̅

𝑍2 =3 as an example, Fig. 4. 4 depicts several calculated
curves based on (4.6) which represent the relationship between̅̅̅̅
𝑍1 , ̅̅̅̅
𝑍2 and the zero-reflection
frequencies. Curves with four different values of ̅̅̅̅
𝑍2 range from 1.5 to 3 are plotted here. If ̅̅̅
𝑍1
̅̅̅2 have only one intersection point then there is only one zero-reflection frequency, and
and 𝑍
̅̅̅1 = 𝑍matched
in this case 𝑍 ̅ . Maintaining ̅̅̅̅
𝑍2 , and slightly decrease̅̅̅̅
𝑍1, two intersection points
will appear, as a result, the approximate matching range is extended. For example, the
intersection points of ̅̅̅̅ 𝑍2 = 2.5 have corresponding frequencies ̅̅̅̅̅
𝑍1 = 0.4 and ̅̅̅̅ 𝑓𝑚1 = 1.4 and
̅̅̅̅̅
𝑓𝑚2 = 4.5 which indicates that ̅̅̅̅̅
𝑓𝑚1 and ̅̅̅̅̅
𝑓𝑚2 are the zero-reflection frequencies over the whole
tuning range. Theoretically, the wider these two frequencies apart, the wider the approximate
matching range can be. When̅̅̅̅
𝑍2 = 1.5, ̅̅̅̅
𝑍1 should employ a value between 0.2 and 0.4 to have
two intersection points on the curve as shown in Fig. 4. 4, otherwise there will be no
approximate matching over the operating band. Although the large gap between these two
frequencies indicates a wide tuning range, such small ̅̅̅̅
𝑍1 (̅̅̅̅
𝑍1 = 0.2~0.4) indicate large width
of the transmission lines which makes it more difficult to connect with the varactors.
Frequency Tuning Range Regarding ̅̅̅̅

𝑪𝟏
This section discusses the relationship between the value of the loaded capacitors̅̅̅̅
𝐶1 and the
corresponding centre operating frequency̅̅̅̅
𝑓𝑚 . The design in [17] has a reflection minimum at
around the original frequency and an additional reflection-zero introduced by the capacitors.
P a g e | 99
Thus, there will be two reflection minima in the passband, and the frequency bandwidth can be
tuned by changing the capacitance of the capacitors. The difference of the proposed work here
is that, by properly choosing̅̅̅̅
𝑍1 , ̅̅̅
𝑍2 , 𝜃1 and 𝜃2 , this design only retains the reflection minimum
introduced by the loaded capacitors. As a result, instead of having a tunable bandwidth as in
[17], this design will have a tunable centre frequency by changing the capacitance.
As the S11 curves in Fig. 4. 5(a)-(d) illustrate, the reflection minimum frequency ̅̅̅
𝑓𝑚 is a
function of ̅̅̅̅
𝐶1 . By choosing an appropriate ̅̅̅
𝐶1 , a minimum |S11| at the chosen frequency can
be generated. Since each S11 curve has only one trough which is the reflection minimum, the
expression of the S11 envelope curve can be calculated based on the derivation of |S11| regarding
frequency. In this case, to find the corresponding capacitance ̅̅̅
𝐶1 and the reflection minimum
frequency ̅̅̅
𝑓𝑚 , the derivative of |S11| should be zero at ̅̅̅
𝑓𝑚 :
d S11
0 (4.13)
d C1
Which gives (4.14).
1
C1  
 2PZ
1
2
  2 4 4 2 2 2
  2
  2 4
 3
 8Y Z1 cos(1 f m ) 2  4  P Z 1  Z 1  4 P Z 1  5Z1 sin(1 f m )  PZ1 Z1  4 cos(1 f m )  P  1 Z1  4 sin(1 f m )  PZ1  4 PZ1
8  2 5  2 
 4  
 1
 4

 4   2 5 2
 f m Z1    P   Z1  2  cos(1 f m )2   4  PZ1  Z12  12  sin(1 f m )  cos(1 f m )  PZ1 Z1  12 sin(1 f m )  2   P   Z1 
1 2
 4 
(4.14)
𝐶1 and̅̅̅̅
The relationship of ̅̅̅̅ 𝑓𝑚 as shown in (4.14) indicates the correlation of the capacitance
range and the frequency tuning range. To further reveal the relationship between |S11| and
reflection minima frequencies ̅̅̅̅
𝑓𝑚 , the envelope of |S11| needs to be derived. As ̅̅̅̅
𝑓𝑚 is a
dependent variable of ̅̅
𝐶̅̅ ̅̅̅̅
1 in (4.13), so 𝐶1 can be replaced by substituting (4.13) to (4.5). Thus,
the envelope of |S11| can be expressed by:
  
2 2
2c  Y Z1 sin(1 f m )  Z1 1  cos(1 f m )  d  Z1 sin(1 f m )

2a  Y Z )  Z 1  cos( f )    3b  Z
S11 envelope 2 2
1 sin(1 f m 1 1 m 1 sin(1 f m )
(4.15)
where a, b, c, d and 𝑌̅ can be found in (4.1) and (4.5).
P a g e | 100
(a) (b)
(c) (d)
Fig. 4. 5 Theoretical S11 envelope curves as a function of 𝑪𝟏 for different impedance of

front transmission lines for (a) α=1, (b) α=1.15, (c) α=0.85 and (d) α=0.6.
By only changing the impedance ̅̅̅

𝑍1 while ̅̅̅
𝑍2 is fixed, the frequency tuning range of PD
changes regularly. Fig. 4. 5(a)-(d) depicts six sampled |S11| curves with different values of ̅̅̅
𝐶1 .
A dash line that is so called ‘envelope curve’, is drawn to illustrate the troughs of |S11| curves
over the whole frequency band. The envelope curve is a group of reflection minima that records
the movement locus of minimum values |S11|min when̅̅̅̅
𝐶1 is changing, and other parameters
remain the same. Let ̅̅̅
𝑍1 = α∙ 𝑍̅matched , or α=1, where α is a rational factor. Fig. 4. 5(a)
represents that when ̅̅̅
𝑍1 = 𝑍̅𝑚𝑎𝑡𝑐ℎ𝑒𝑑 , there is only one zero-reflection point over the tuning
P a g e | 101
range. In this case, the centre frequency tuning range is relatively narrow. If α increases, for
instance, α=1.15, the reflection-zero disappears and the minimum of the S11 envelope will be
worse, and the tuning range will shrink as shown in Fig. 4. 5 (b). Conversely, there will be two
reflection-zeros on the S11 envelope curve if α decreases from 1 (e.g. α=0.85, 0.6) as shown in
Fig. 4. 5(c) (d). In this case, the frequency tuning range can be extended by controlling the
distance between the two reflection-zeros. A larger distance between the two minima leads to
a wider frequency tuning range. However, it should be noted that the S11 between the two zeros
will be sacrificed.
Tuning range of the tunable PD
Fig. 4. 6 The relationship between the worst in-band S11 and changing ̅̅̅
𝐙𝟏 .
The worst in-band |S11| is the other crucial factor that needs to be considered for bandwidth
calculations. Although large difference between the two reflection-zeros indicates a wide
tuning range, the S11 envelope should be better than a specified value. For example, if S11 is
expected to be better than 20 dB, α should be greater than 0.6 because |S11| at the centre
frequency will exceed 20 dB as shown in Fig. 4. 5(d) if α is smaller than 0.6, which does not
meet the design specification. Mathematically, the worst in-band S11 can be calculated by
P a g e | 102
taking the derivative of the S11 envelope. Due to the complexity of the calculation, Fig. 4. 6
can be used to give a picture of how the worst in-band S11 varies with a changing ̅̅̅
𝑍1 for
different ̅̅̅̅
𝑍2 . The curves in Fig. 4. 6 are plotted in Matlab based on four sets of calculations.
̅̅̅2 (𝑍
By substituting 𝜃1 = 30𝑜 , 𝜃2 = 18𝑜 , and 𝑍 ̅̅̅2 = 1.5, 2, 2.5 and 3) into (4.15), the envelope
of |S11| can be expressed as a function of the reflection minimum frequency ̅̅̅ ̅̅̅1 .
𝑓𝑚 and 𝑍
̅̅̅1 from 0.2 to 0.8 in Matlab, the corresponding worst in-band S11 can then be
Sweeping 𝑍
obtained using the derivation of |S11|envelope regarding ̅̅̅
𝑓𝑚 . For a fixed̅̅̅̅
𝑍2 , the smaller ̅̅̅
𝑍1 is, the
higher the in-band S11 will be, because a smaller ̅̅̅
𝑍1 results in a wider frequency difference
between the two reflection zeros. It can also be observed that a smaller ̅̅̅̅
𝑍2 will result in a wider
frequency difference with a fixed ̅̅̅
𝑍1 . Since 20 dB is often the desired in-band S11 value in PD
design, a vertical marker line is drawn and the four interception points are̅̅̅̅ ̅̅̅2 = 1.5,
𝑍1 = 0.3, 𝑍
̅̅̅ ̅̅̅2 = 2, ̅̅̅
𝑍1 = 0.38, 𝑍 𝑍1 = 0.45, ̅̅̅
𝑍2 = 2.5, and ̅̅̅ ̅̅̅2 = 3 , respectively. With these
𝑍1 = 0.5, 𝑍
impedance combinations, the worst in-band value of the |S11| envelope is exactly 20 dB, to
make full use of the tuning band. Nevertheless, the frequency tuning ranges for these four sets
of impedances are different, the largest tuning range in this case is realized when ̅̅̅
𝑍1 =
̅̅̅2 = 1.5 referring to the previous discussion.
0.3, 𝑍
Another factor that affects the frequency tuning range of the PD is the capacitance tuning range
of the varactors. Fig. 4. 7 shows an example of the tuning range. The two curves imply that the
frequency tuning range of the PD varies as a function of the required capacitance for varactor̅̅̅̅
𝐶1
and ̅̅̅̅
𝐶2 . The blue curve exhibits the tuning range of ̅̅̅̅
𝐶1 . The capacitance versus reflection
minimum frequency can be solved using Matlab by substituting ̅̅̅̅̅
𝑍1 = 0.4, ̅̅̅̅̅
𝑍2 = 2, and 𝜃1 = 30°,
𝜃2 = 18° to (4.14). Once the tuning range of ̅̅̅̅
𝐶1 is determined, substituting ̅̅̅̅
𝐶1 together
with̅̅̅̅̅̅
𝑍1,𝑍 ̅̅̅̅̅̅ ̅̅̅̅
2 , 𝜃1 and 𝜃2 to (4.11) will give the relationship between 𝑍𝐼𝑆𝑂 and 𝑓𝑚 . According to
̅̅̅̅̅
(4.12), the tuning range of ̅̅̅̅

𝐶2 can be worked out by extracting the imaginary part of ̅̅̅̅̅̅
𝑍𝐼𝑆𝑂 . By
specifying |𝑆11 |, the specified maximum operating frequency ̅̅̅̅̅̅
𝑓max and the minimum operating
̅̅̅̅̅
frequency𝑓 ̅̅̅̅̅̅̅ ̅̅̅̅̅̅̅
min , the corresponding 𝐶1min and 𝐶1max can be found using (4.12) and (4.13). To
simplify the design process, ̅̅̅̅̅̅̅ 𝐶1max can be found from Fig. 4. 7 by specifying ̅̅̅̅̅̅
𝐶1min and ̅̅̅̅̅̅̅ 𝑓max
and ̅̅̅̅̅
𝑓min ,, then |𝑆11 | will be evaluated if it is better than the specified value at ̅̅̅̅̅̅
𝑓max and ̅̅̅̅̅
𝑓min ,.
The frequency tuning range depends on the tuning range of the varactors while it is usually
limited by the mechanism of the varactor diodes. Although a wide frequency tunable range can
be achieved theoretically, the range will be limited by the capacitance range of the varactors in
P a g e | 103
practice. In that case, the required ̅̅̅̅̅̅̅ 𝐶1max for ̅̅̅̅̅̅

𝐶1min and ̅̅̅̅̅̅̅ 𝑓max and ̅̅̅̅̅
𝑓min , should be in the
capacitance tuning rage of the commercial varactors or as much as possible.
Fig. 4. 7 Theoretical required capacitance ̅̅̅

𝑪𝟏 and ̅̅̅
𝑪𝟐 versus corresponding frequency
with ̅̅̅
𝐙𝟐 =2 when ̅̅̅
𝐙𝟏 =0.4, 𝜽𝟏 = 𝟑𝟎° and 𝜽𝟐 = 𝟏𝟖°.
Design Guidance
The design procedure of the proposed tunable PD can be summarized as follows.
1) Specify the desired frequency range ̅̅̅̅̅

𝑓min and ̅̅̅̅̅̅
𝑓max .
2) Choose appropriate 𝜃1 and 𝜃2 based on the desired frequency range. Determine ̅̅̅
𝑍1 and ̅̅̅
𝑍2
referring to (4.6) and Fig. 4. 4, making sure the chosen frequency range is slightly wider than
the two reflection-zero frequencies. Besides, ensure that the worst in-band S11 would meet the
specification.
3) Verify if |𝑆11 | at ̅̅̅̅̅̅

𝑓max and ̅̅̅̅̅
𝑓min would satisfy the specification based on (4.14) and (4.15). If
not, choose other values for 𝜃1 and 𝜃2 and repeat design procedure from step 2).
4) Calculate the required capacitance range for varactor ̅̅̅

𝐶1 .
P a g e | 104
5) The resistance 𝑅̅ and capacitance ̅̅̅

𝐶2 on the isolation circuit can be calculated by substituting
all previous design parameters into (4.11) and (4.12). The required tuning range of ̅̅̅
𝐶2 on the
isolation circuit can be estimated if ̅̅̅
𝐶1 and the corresponding frequency f are specified.
6) Choose varactors that can achieve the required tuning ranges of ̅̅̅
𝐶1 and ̅̅̅
𝐶2 .
7) After obtaining these parameters, implement the PD with distributed elements and optimize
the physical dimensions with the aid of EM simulation tools. Bias circuits will also be added
to tune the varactors in the measurement.
4.4 Fabrication and Measurement
(a)
(b)
Fig. 4. 8 (a) Configuration of the proposed PD and (b) photo of the fabricated PD.
P a g e | 105
Table 4. 1 DESIGN PARAMETERS OF THE PROTOTYPE TUNABLE POWER

DIVIDER.
𝑊1 (mm) 4.3 C1 (pF) 5.1-0.01

𝑊2 (mm) 2.3 C2 (pF) 2.1-0.2
𝑊3 (mm) 0.5 R (Ω) 100
𝐿1 (mm) 21 V1 (V) 1.7-20
𝐿2 (mm) 7 V2 (V) 0-20
𝐿3 (mm) 12
One prototype tunable PD with a frequency tuning range from 1 GHz to 4.2 GHz was designed
and fabricated for the validation of the proposed method. Based on the desired tuning range,
the electric lengths of the front and isolation transmission lines are 𝜃1 = 30° and 𝜃2 = 18° at 1
GHz. According to (4.6), ̅̅̅ ̅̅̅1 = 2.4 are selected in the design. Hence the
𝑍1 = 0.4 and 𝑍
characteristic impedances are 𝑍1 = 20.3 Ω and 𝑍2 = 120 Ω. The corresponding |S11| at 1 GHz
and 4.2 GHz are both lower than -20 dB. Then the required capacitance range for 𝐶1 is obtained
to be 0-7.9 pF. Theoretically, as shown in Fig. 4. 5(d), the achievable frequency tuning range
is 4.7:1. For the design parameters in isolation circuit, the resistor R is chosen as 100 Ω
according to (4.11) and (4.12), and the tuning range of 𝐶2 is 0.2 pF to 2 pF. Based on the desired
capacitance tuning ranges of 𝐶1 and 𝐶2 , surface mount tuning varactors MA46H 202 (7-0.5 pF)
are chosen for 𝐶1 , while 𝐶2 is realized by an SMV 2202 (2.1-0.23 pF). Similar to the process
in Chapter 3 Section 3.5.1, the calculated parameters are converted to distributed elements by
using Sonnet simulation. The substrate used in Chapter 3 is employed in this design. Fig. 4. 8
(a) shows the configuration of the design with bias circuits supplying voltages to the varactors.
Proper gaps are reserved on the layout for soldering lumped elements. Optimized design
parameters are listed in Table 4.1. Design layout file is imported to LPKF CircuitCAM for the
fabrication process. The fabricated circuit is shown in Fig. 4. 8(b). The front transmission lines
are bent in order to connect the varactors (MA46H-202) in parallel with them. Resistor R and
a capacitor (SMV-2202) connecting in series are soldered to the isolation circuit. There are
three DC bias circuits added to the divider. 20 pF capacitors and 10 kΩ resistors are used for
DC blocking and RF blocking purposes in the bias circuits, respectively.
A DC power supply (BK PRECISION 9202) and a VNA (Agilent FieldFox N9917A) are
employed in the measurement. The varactor diodes are operated under reverse bias conditions,
so that there is no conduction. The capacitance 𝐶1 is increased from 0.5 pF to 5.1 pF as the bias
P a g e | 106
voltage for the varactors in the main transmission lines is decreased from 20V to 0V.
Conversely, the capacitance 𝐶2 is decreased from 2.1 pF to 0.2 pF as the bias voltage on the
isolation circuit varies from 0 to 20 V. Since the capacitance range of these varactors cannot
cover the required tuning range of 𝐶1 and 𝐶2 , two capacitors with fixed values of 0.01 pF and
0.2 pF are used in turn here to replace the varactors for part of the measurements. In the
measurement, gradually tune the bias voltage on MA46H 202 to 1.7V, 6.1V, 11.4V and 20V
to give S11 band located at 0.9 GHz, 1.7 GHz, 2.1 GHz, 2.8 GHz, respectively. Besides, 0.01
pF and 0.2 pF capacitors are used in turns to replace MA46H-202 to tune the centre frequency
to 3.45 GHz and 4.2 GHz. At the same time, in order to make sure S22 and S23 move along with
S11, the bias voltage on SMV-2202 is set to be 0V, 0V, 0V, 2.2V, 7.8V and 20V, respectively.
The measurement procedure is repeated several times, and the precision of the measurement is
±0.1dB.
Fig. 4. 9 and Fig. 4. 10 show the comparison of the simulated and measured S-parameter
responses when the divider is tuned to the six sampled frequency bands accordingly. The
measured S11 and S22 are below -20 dB at each band. The insertion loss varies between 3.2 dB
and 4 dB, or 0.2 dB to 1 dB above the intrinsic loss of 3 dB. The measured isolation S23 is better
than -20 dB at most of the sampled frequencies. The isolation is slightly worse at 4.2 GHz with
a value of -17 dB, which is probably caused by fabrication errors and the accuracy of 𝐶2 . The
measured tunable frequency range is 0.9 GHz to 4.2 GHz. The tuning range is 4.67:1 as the
measured results indicate. Overall, the measured results are in very good agreement with the
simulated ones. To show the advantages of the proposed PD, Table 4.2 lists the performance
comparison between this work and other works reported in recent years. As the table illustrated,
the frequency tuning range (when S11<-20 dB) for the proposed work is much wider than other
relevant works. The measured in-band S11 (better than 25 dB) is better than all other works.
The circuit is very compact as well for such a wide tuning range.
P a g e | 107
(a)
(b)
Fig. 4. 9 Comparison between simulations and measurements, (a) S11 and (b) S21.
P a g e | 108
(a)
(b)
Fig. 4. 10 Comparison between simulations and measurements, (a) S22, and (b) S23.
P a g e | 109
Table 4. 2 COMPARISON TO OTHER WORKS.
Ref. Tuning Range (GHz) Return Loss (dB) Insertion Loss (dB)
[10] 1.3-2.08 (1.6:1) ≥ 20 5.3-2.9

[11] 0.62-0.85 (1.37:1) ≥ 15 5.4-4.8
[13] 0.56-1.39 (2.48:1) ≥ 22 4.5
[14] 0.5-1.3 (2.6:1) ≥ 17 4.6-3.5
[15] 0.83-2.4 (2.9:1) > 23 3.5-3.3
[16] 0.85-2.4 (2.82:1) > 20 5.6-2.16
This work 0.9-4.2 (4.67:1) ≥ 25 4-3.2
4.5 Summary
A compact PD with a wide frequency tuning range has been designed and verified in this
chapter. The proposed structure is based on a Horst-type PD, and a pair of varactors is added
in parallel with the main transmission lines which introduces a controllable reflection minimum
sensitive to the capacitance change. A single varactor diode is added in series to the isolation
circuit to tune the operating bands of output return losses and isolation. The design equations
have been derived to quantificate the changing center frequency with capacitances. A design
guidance is provided to determine the parameters of the proposed PD. Impedances and the
tuning range of the varactors should be carefully selected for achieving desired tuning
bandwidth. The centre frequency tuning range is dramatically broadened by using the proposed
method. To validate the proposed theory, a prototype with a frequency tuning range from 0.9
GHz to 4.2 GHz has been fabricated and measured. In the measurements, two sets of bias
circuits are added to the design to alter the capacitance of the varactors. To the best of authors’
knowledge, the tuning range is superior to all other works. The circuit has a relatively very
compact size (0.2 λg × 0.16 λg). The measured and simulated results have a perfect agreement
with each other. Moreover, this method can also be applied to multi-way designs. The reason
is that the operating band of the two-way PD exhibits a very good response but within a narrow
bandwidth. When several two-way PDs are cascaded together, the impedance matching at the
interconnecting points will be good as well, so that the overall circuit will obtain the same
frequency tuning range as the two-way PD exhibits.
P a g e | 110
References
[1] X.-D. Wang, J.-P. Wang, G. Zhang, J.-S. Hong and W. Wu, "Dual-Wideband Filtering Power
Divider With Good Isolation and High Selectivity," IEEE Microwave and Wireless Components
Lett., vol. 27, no. 12, pp. 1071-1073, 2017.
[2] D. Psychogiou, R. Gómez-García, A. C. Guyette, and D. Peroulis, “Reconfigurable single-/multi-
band filtering power divider based on quasi-bandpass sections,” IEEE Microw. Wireless Compon.
Lett., vol. 26, no. 3, pp. 684–686, Sep. 2016.
[3] Wu, Y., Liu, Y., Xue, Q., Li, S., and Yu, C.: ‘Analytical design method of multi-way dual-band
planar power dividers with arbitrary power division’, IEEE Trans. Microw. Theory Tech., 58, (12),
pp. 3832–3841, 2010.
[4] Y.-F Pan, S.-Y Zheng, Y.-M. Pan, Y.-X. Li and Y.-L. Long, “A Frequency Tunable Quadrature
Coupler with Wide Tuning Range of Center Frequency and Wide Operating Bandwidth,” IEEE
Trans. Circuits Syst. II, Exp. Briefs, vol. 65, no. 7, pp.864-868, 2018.
[5] T. Yang, and G. M. Rebriel, “Bandpass-to-Bandstop Reconfigurable Tunable Filters with
Frequency and Bandwidth Controls,” IEEE Trans. Microw. Theory Techn., vol. 65, no. 7, pp.
2288-2297, 2017.
[6] Z. Wang, J. R. Kelly, P. S. Hall, A. L. Borja, and P. Gardner, “Reconfigurable parallel coupled
band notch resonator with wide tuning range,” IEEE Trans. Ind. Electron., vol. 61, no. 11, pp.
6316–6326, Nov. 2014.
[7] Zhi-Han Chen and Qing-Xin Chu: 'Dual-band reconfigurable bandpass filter with independently
controlled passbands and constant absolute bandwidths', Microw. Wireless Compon. Lett., 2016,
26, (2),pp.92-94.
[8] L. Athukorala, D. Budimir, "Open-loop tunable resonators and filters with constant bandwidth,"
Microwaves, Antennas & Propagation, IET, vol.6, no.7, pp.800-806, May 16, 2012.
[9] H. Zhu, and A. M. Abbosh, “Tunable balanced bandpass filter with wide tuning range of center
frequency and bandwidth using compact coupledline resonator,” IEEE Microw. Wireless Compon.
Lett., vol. 26, no. 1, pp. 7--9, Jan. 2016.
[10] P.-L. Chi and T. Yang, “A 1.3–2.08 GHz Filtering Power Divider With bandwidth control and
high in-band isolation,” IEEE Microw. Wireles Compon. Lett., vol. 26, no. 6, pp. 407-409, 2016.
[11] L. Gao, X. Y. Zhang, and Q. Xue, “Compact Tunable Filtering Power Divider With Constant
Absolute Bandiwdth,” IEEE Trans. Microw. Theory Techn., vol. 63, no. 10, pp. 3505-3513, 2015.
[12] A.-L. Perrier, O. Exshaw, J.-M. Duchamp, and P. Ferrari, “A Semi-Lumped Miniaturized Spurious
Less Frequency Tunable Three-port Divider\combiner with 20 dB Isolation Between Output Ports,”
in IEEE MTT-S International Microwave Symposium Digest, San Francisco, 2006.
P a g e | 111
[13] X. Shen, Y. Wu, S. Zhou, and Y. Liu, “A novel coupled-line tunable wikinson power divider with
perfect port match and isolation in wide frequency tuning range,” IEEE Trans. Compon., Packag.
Manuf. Technol. , vol. 6, no. 6, pp. 917-925, 2016.
[14] S. Lin, Y. Chen, P. Chiou and S. Chang, "Tunable Wilkinson power divider utilizing parallel-
coupled-line-based phase shifters," IEEE Microw Wirel Components Lett 27(4), pp. 335–337, 2017.
[15] T. Zhang and W. Che, “A Compact Tunable Power Divider With Wide Tuning Frequency Range
and Good Reconfigurable Responses,” IEEE Trans. Circuits Syst. II, Exp. Briefs, vol. 63, no. 11,
pp. 1054-1058, 2016.
[16] T. Zhang, X. Wang and W. Che, “A Varactor Based Frequency-Tunable Power Divider With
Unequal Power Dividing Ratio,” Microwave and Wireless Components Letters, IEEE, vol. 26, no.
8, pp. 589-591, 2016.
[17] A. Chen, Y. Zhuang, Y. Huang and J. Zhou, ‘A Horst-type Power Divider with Wide Frequency
Tuning Range Using Varactors,’ IEEE Access, ISSN: 2169-3536， pp: 1-13, 2018.
[18] S. Horst, R. Bairavasubramanian, M. M. Tentzeris, and J. Papapolumerou, “Modified Wilkinson
power dividers for millimeter wave applications,” IEEE Trans. Microw, Theory Techn., vol. 55,
no. 11, pp. 2439-2446, 2007.
[19] X. Wang, I. Sakagami, N. Ito, and A. Mase, “Miniaturized Horst-type Wilkinson power divider
with simple layout,” Electron. Lett., vol. 49, no. 6, pp. 384-395, 2013.
P a g e | 112
Chapter 5: Millimetre-Wave CPW PDs Using Spiral Resonators for Miniaturization.
Chapter 5 Millimetre-Wave CPW PDs Using

Spiral Resonators for Miniaturization.
5.1 Introduction
Fig. 5. 1 Relative loss, size, and cost of various RF resonators [1].
PD designs for mm-Wave applications require low loss, compact size, low cost and easy to be
integrated into MMICs [1]. Fig. 5. 1 compares these features between various technologies
including superconductors, microstrip/striplines, substrate integrated waveguide (SIW),
dielectric resonators and conventional metal waveguides [1]. For such high frequencies, the
traditional waveguide is mostly used due to its high Q and low power loss. However, the
traditional waveguide suffers from its massive size which is not preferred for circuit integration.
To overcome this problem, SIW technique has been introduced. The substrate with via-holes
can transmit signal just as in a waveguide with a much higher permittivity compared to the air
used in the conventional waveguides, leading to a compact circuit size [2]. However, the
substrate used in SIW operating in mm-Wave will be relatively lossy [1][3][4].
P a g e | 113
Superconducting components can be applied to achieve both a compact size and sufficient low
loss. However, the cost is too high to perform mass production. Therefore, to make a trade-off
among these features, coplanar PDs are proposed using coupled resonators.
The coplanar waveguide (CPW) was firstly proposed by Wen [5]. It consists of a metal strip
and two adjacent ground planes printed on the dielectric substrate. The most significant
advantage of the CPW is that only a single-layer metal process is required. As via-hole
grounding can be omitted because the ground and metal strip are uniplanar so that the active
or passive components in shunt configuration can be easily connected. On the other hand, the
conducting material used in the CPW structure is still relatively lossy in mm-Wave [6]-[9].
Typically, a three-port component requires a T-junction to connect two output ports with the
input to form a PD. These connections occupy larger sizes and produce more conducting loss
especially for mm-wave applications. Moreover, the unwanted quasi-TEM mode will rise due
to the intrinsic structure of CPW [3]. Air-bridges are needed to suppress the undesired mode
by keeping the lateral ground planes at the same potential, which will indeed complicate the
fabrication process and introduce more loss [4]. Many CPW-PD designs have been reported
[10][11] recently. They used long CPW lines (quarter wavelength) that require more air-bridges
to maintain the desired operating mode, resulting in large sizes and losses. To overcome the
problems discussed above, [12]-[16] introduced CPW coupled resonators. [16] utilises three
stepped resonators in the design to eliminate air-bridges. However, this type of resonators is
not easy to be implemented at mm-Wave, and the topology is complicated to be used for multi-
way PDs.
The spiral resonators have been adopted in mm-Wave filters designs which can be realised in
the forms of CPWs [17] [18]. The spiral resonators can achieve a very compact circuit size by
folding a quarter-wavelength CPW central strip into a spiral. Referring to the size comparison
between the spiral resonator and open-loop resonator in [19], the spiral resonator only occupies
40% of the area of an open-loop at the same resonant frequency. The spiral resonators are not
only compact but also have remarkable properties such as high Q, high power capability and
insensitivity to the fabrication process [9]. Hence, the CPW coupled resonators are firstly
adopted in mm-Wave PD designs. In this work, the conducting loss can be minimised by
replacing the interconnecting lines with coupled resonators. G type spiral resonators are utilised
P a g e | 114
in this design for size miniaturisation. Moreover, the Q can also be improved which will be
explained in detail in the following sections.
5.2 Analysis of the Coupled-Resonator PD
To design a low loss and compact size PD, three CPW spiral resonators are employed in the
circuit. To transmit power from one input port to two output ports, an equivalent circuit of three
coupled resonators is depicted in Fig. 5. 2.
Fig. 5. 2 The equivalent circuit of a PD with three coupled resonators.
The inductance, capacitance and resistance of each resonator are denoted as L, C, and R in Fig.
5. 2, respectively. I represents the loop current in each resonator while Vs indicates the source
voltage. M12 and M13 represent the mutual inductances between Resonator 1 and 2, and
Resonator 1 and 3, respectively. By using the voltage law, the algebraic sum of the voltage
drops around any closed loop circuit in a network is zero. Based on the Kirchhoff’s Voltage
Law (KVL), the voltages on the resonators can be expressed as:
P a g e | 115
 1 
 1
R  j L   I1  j M12 I 2  j M13 I 3  VS (5.1)

1
j C1 
 1 
 j M12 I1   R2  j L2   I 2  j M 23 I 3  0 (5.2)
 jC2 
 1 
 j M13 I1  j M 23 I 2   R3  j L3   I3  0 (5.3)
 j C3 
It can be assumed that the coupling between the two outputs can be ignored as it is very weak
compared to the main coupling from the input port, yielding M23 = 0.
The port termination of each port is usually the same at 50 Ω, therefore, that is, R1 = R2 = R3
= R = 50 Ω. Then, when the system is operating at the resonant frequency ω0, the current flow
in each loop can be derived as:
RVS
I1  (5.4)
( R 2  02 M 122  02 M 132 )
j M 1nVS
In  , (n  2or 3) (5.5)
( R  02 M 122  02 M132 )
2
Now, the scattering parameters of the system can be evaluated as:
2 RI1 2R2
S11  1   1 2 (5.6)
VS ( R  02 M122  02 M132 )
2 RI n 2 j M 1n R
Sn1   2 , (n  2or 3) (5.7)
VS ( R  02 M 122  02 M 132 )
It can be found that the power splitting ratio of the output ports can be derived as:
S21 / S31  M12 / M13 (5.8)
P a g e | 116
On the other hand, to minimise the reflection coefficient to enable the maximum power to be
injected to the system, S11 should be equal to zero, yielding:
2R2
1 2 0
( R  02 M 122  02 M 132 )
R 2  02 ( M 122  M 132 )  0
(5.9)
R
M M2 2

12 13
0
Then, by combining the desired power splitting ratio derived in (5.8) (5.9), the desired mutual
inductances to realise the PD can be obtained. The same method can be applied to design an
N-way PD with desired power division ratio.
Coupling coefficient kij is usually employed to describe the relationship between the distances
of resonators and the coupling strength, while mutual inductance M1n cannot show intuitive
relationship between distance and coupling strength. So introducing k1n to equations can
simplify the realisation of designs. M1n and k1n can be expressed as:
M1n
k1n  (5.10)
L1 Ln
And introducing external quality factor Qe into the equation,
1 R
 n (5.11)
Qen 0 Ln
where 𝜔0 = 1⁄√𝐿𝐶 .
So the power dividing ratio of the output ports can be concluded as
k12 Qe 2
S21 / S31   (5.12)
k13 Qe3
P a g e | 117
As the Eq (5.12) indicates, the power division are not only relates to the coupling coefficients,
but also affected by the ratio of their external quality factors.
5.3 CPW PD with Spiral Resonators
Fig. 5. 3 The configuration of the proposed CPW PD with three resonators.
To reduce the circuit footprint, spiral resonators are employed for significant size reduction
compared to conventional ring resonators. To verify the method, a CPW-PD with three spiral
resonators is designed with a centre frequency of 48.5 GHz. The configuration of the design is
shown in Fig. 5. 3. The spiral line winds clockwise with one end connected to the ground. The
length of the resonator is about λ⁄4. The width W and gap S of the spiral line are both chosen
to be 10 μm while the distance between two adjacent resonators is denoted as d. All the
resonators are directly connected with the input 50 Ω CPW line where Sport is 15μm and Wport
is 20μm. Due to the winding shape of the spiral resonator, if two resonators are coupled to the
common resonator at two sides, these three resonators cannot be centrosymmetric. As a result,
the common and upper resonators are wound in one direction while the lower one in the
opposite direction to the common resonator winding.
P a g e | 118
Resonant Frequency
Unit: um
(a)
(b)
Fig. 5. 4 (a) The configuration of a quarter-wavelength spiral resonator, and (b) the
relationship between resonance frequencies and side-length L.
To examine the resonant frequency f0, a pair of open CPW feed-lines are placed at both sides
of the resonator with some distance. The side-line length of the spiral resonator l is chosen to
be 110 μm, 112 μm, and 115μm, and the overall length of the spiral line is 596 μm, 614 μm
and 656 μm, respectively. Fig. 5. 4(a) shows the configuration of a single spiral resonator for
P a g e | 119
computation and Fig. 5. 4(b) depicts the variation of resonance frequency f with side-length l
of the spiral. In obtaining the relationship given in Fig. 5. 4(b), the physical dimensions in Fig.
5. 4(a) are applied. The curve illustrates that the resonant frequency in general increases as the
length of the spiral line decreases. In this chapter, a PD with a centre frequency f0 equals to
48.5 GHz is designed. As a result, l is chosen to be 115 μm to conduct the following designs.
Coupling Coefficient k
Unit: um Unit: um
(a) (b)
Fig. 5. 5 (a) Face-to-back oriented, and (b) back-to-back reversely oriented resonators.
Fig. 5. 5 illustrates two types of coupled spiral resonators, one is face-to-back orientation and
the other in back-to-back direction. The distances between every two resonators are denoted as
d1 and d2 respectively. Simulation of the coupling can be carried out by placing a pair of feed
lines to the two resonators in the opposite direction. With the coupling, the resonant frequency
will be split into two, frequencies f1 and f2 as shown in Fig. 5. 6(a)-(b). The inter-resonator
coupling coefficient can be calculated by
𝑓2 2 −𝑓1 2
𝑘= (5.13)
𝑓2 2 +𝑓1 2
P a g e | 120
(a) (b)
(c)
Fig. 5. 6 Resonant mode splitting phenomena of (a) face-to-back oriented and (b) back-
to-back reversely oriented. (c) Coupling coefficient k versus the distance d between
resonators.
Generally, with the increase of distance d, the split resonance frequencies f1 and f2 come closer,
which means the coupling is weaker. The coupling coefficient k is computed, and the variation
with different distance d is shown in Fig. 5. 6(c).
P a g e | 121
External Coupling Qe
The general behaviour of these coupled resonators is affected by parameters, including the
position and offset of the feeding lines, the gaps between spiral line and the ground and the
height of the air gap. To achieve optimal performance, a parametric study on realizing external
Qe is performed in this section. For the tapped-line coupling, a 50 Ω CPW line is directly tapped
onto the input and output resonators. Lfeed denotes the distance between the point where the
spiral line is connected to the ground and the CPW feed-point. The external single loaded
quality factor can be extracted from simulation by using:
𝜔0
𝑄𝑒 = (5.14)
∆𝜔±90°
Where 𝜔0 = 1⁄√𝐿𝐶 is the resonant frequency, and ∆ω±90° represents the absolute bandwidth
between the ±90° points which should be determined from the frequency at which the phase
shifts ±90° with respect to the absolute phase at 𝜔0 .
Unit: um
(a) (b)
Fig. 5. 7 (a) Feeding structure of the PD, and (b) relationship between external Qe and
centre frequency f0 with the off-set Lfeed in (a).
Fig. 5. 7 (b) illustrates that the loaded quality factor is decreasing with the distance between
the spiral-line end to the feed line increases. The Qe achieves the maximum value when the
feed-line is directly connected to the spiral line end. This is because the smaller the Lfeed is, the
closer is the tapped line to a virtual grounding of the resonator, which results in a weaker
coupling, in other words, a larger external Qe.
P a g e | 122
Unit: um
(a) (b)
Fig. 5. 8 (a) Feeding structure of the resonator, and (b) variation of Q e and centre
frequency f0 with the gap width in (a).
Another way to increase the external coupling is to change the gap width G between the spiral
and the ground on the feeding side. As is shown in Fig. 5. 8(b), when the gap width G is
extended, the external coupling is enhanced significantly. However, extending the gap width
G will also result in a centre frequency shifting down. To increase the external coupling without
changing the centre frequency, the inner spiral line can be shorted a little bit to compensate. To
get higher or lower values for the external quality factor Qe is a matter of choosing the proper
side for feeding the resonator.
5.4 Configurations and Designs
Proposed PDs
Fig. 5. 9(a) shows the structure of the proposed CPW two-way PD. The feedings at all three
ports have an offset instead of feeding in the middle of the first spiral line to enhance external
coupling. The signal is passing through the common resonator and coupled to upper and lower
resonators with the same amount of the coupling. That is, the coupling coefficients 𝑘12 and 𝑘13
are the same. When the distance between the lower resonator and the common resonator
P a g e | 123
(a) (b)
(c) (d)
Fig. 5. 9 (a) Configuration of the optimized three-resonator PD with a 1:4 power division
ratio, (b) 1:1 power division, (c) 1:2 power division and (d) 1:4 power division.
increases, the coupling strength between them will decrease, less energy will be coupled to the
lower coupling path (from Port 1 to Port 3) and vice versa. It is the same for Port 1 and Port 2.
For equal splitting, the mutual coupling coefficients be 𝑘12 = 𝑘13 = 0.026, and 𝑄𝑒2 = 𝑄𝑒3 =
38.43 to achieve a 3-dB bandwidth of 2.2 GHz at a centre frequency of 48.5 GHz. Then the
corresponding physical distance between coupled resonators can be read from Fig. 5. 9(c). Also,
this coupling structure is required to have a coupling coefficient ratio of 𝑘12 /𝑘13 = 1:2 and 1:4
according to (5.8) for unequal power splitting.
P a g e | 124
(a)
(b)
Fig. 5. 10 (a) Configuration of the CPW PD with five resonators, and (b) its simulated
performance.
A PD with five asynchronous spiral resonators is shown in Fig. 5. 10(a). More spiral resonators
are used to obtain a sharp-selectivity bandpass response compared to the third-order PD for the
same Q𝑒 . However, compared to the three-resonator PD, the power division ratio is more
difficult to control due to the introduction of more resonators. These resonators are not
allocated in one axis for the purpose of circuit miniaturization. The coupling path requires
resonators with different orientations which would severely change the coupling conditions
P a g e | 125
between resonators because of the asymmetry of the spiral structure. As Fig. 5. 10(b) indicates,
the bandwidth is around 1.13:1, and the coupling between Resonator 1 and Resonator 3 is
stronger than the other that of Resonator 1 and Resonator 2 which results in a power division
around 1:8.
Splitting Frequencies (diplexer)
Similar to a PD, a diplexer is also a three-port network, but assign disparate frequency bands
at the same time. In this section, a five-resonator diplexer is designed by using resonators of
different sizes and controlling the coupling strengths between them.
Fig. 5. 11 RF front end of a cellular base station [20].
A diplexer is commonly used in satellite communication systems to combine both the Tx

(Transmit) and the Rx (Receive) antennas on spacecraft. It is frequently used in the RF front
end of cellular radio base stations to separate Tx and the Rx channels as shown in Fig. 5. 11.
The response of a diplexer contains two individual frequency bands, the lower passband and
the higher passband. Conventionally, diplexers can be formed by connecting two separately
designed filters via an external energy distribution device, such as a T-junction [20][21], a Y-
junction [22], a circulator [23] or a common resonator [24][25]. The external junctions utilised
in the conventional approaches for diplexer design resulte in more complex and larger size
devices. Also, the connecting device does not contribute to the resultant diplexer but increases
circuit size. To overcome these drawbacks, referring to designs [16][24][25], the work in this
P a g e | 126
section eliminates the external junctions and replacing it with a common resonator to distribute
the input signal.
Fig. 5. 12 The topology of the proposed resonator-based diplexer.
Fig. 5. 12 shows the topology of the proposed 5-resonator diplexer with each circle represents
a resonator and the interconnecting lines represent the coupling between them. The diplexer
can be divided into two parts. The leading part is directly connected to the input which is called
the stem. The other part is the two branches, one including Resonator 2 and 3 and the other
formed of Resonator 4 and 5. Resonator 2 and 4 are coupled to the common Resonator 1. The
common resonator or Resonators 1 has dual-resonant response to attenuate signals outside of
the lower and higher passbands [26]. The two branches (B and C) work like two individual
single band bandpass filters. Each branch occupies one of the two passbands of the dual-band
filter so that the input signal passing through the common resonator will be guided to one of
the branches, but rejected by the other branch. The general response of the device is given in
as well as the lower and higher frequency bands for the two bandpass filters. 𝑓𝑐 denotes the
centre frequency of the whole circuit. As shown in Fig. 5. 13, the passbands |𝑓2 − 𝑓1 | and |𝑓4 −
𝑓3 | are not only the responses of the two branches, respectively, but also the passbands for the
dual-band filter. Let us define A, B, and C as the common structure, and two passband filters,
respectively. 𝐴−𝑞𝑒1 , 𝐵−𝑞𝑒1 and 𝐶−𝑞𝑒1 in Fig. 5. 12 represent the external coupling between the
ports to the terminal resonators for the filters. 𝐵−𝑚12 and 𝐵−𝑚23 , and 𝐶−𝑚12 and
𝐶−𝑚23 represent the internal coupling for the bandpass filters which can be directly employed
as the internal coupling of the diplexer.
P a g e | 127
(a)
(b)
Fig. 5. 13 Theoretical response of (a) the dual-band bandpass filter, and (b) the bandpass
filters of lower and higher frequency bands, respectively [22].
Normally, the coupling between the common resonators is strong, as well as the coupling
between Resonator 2 and 3, and Resonator 4 and 5. On the contrary, the coupling between the
branches and the common resonator, for example, Resonator 1 and 2 is weak. Moreover, the
common resonator can always be replaced by any networks that have dual-band responses.
For a diplexer with asynchronously frequency bands, the sizes of the resonators on the two
branches (1-2-3 and 1-4-5) are different, so that the self-coupling 𝑚𝑖,𝑖 is introduced here. The
self-coupling 𝑚𝑖,𝑖 is used to quantify the difference between the resonant frequency 𝑓𝑖 of
Resonator i and the canter frequency 𝑓𝑐 of the whole circuit. When resonator i is synchronously
P a g e | 128
tuned, which normally means the size of the resonators are the same, the normalized self-
coupling 𝑚𝑖,𝑖 equals to zero, while in asynchronously tuning, 𝑚𝑖,𝑖 is an non-zero term.
According to the coupling analysis in [26], the relationship between the normalized self-
coupling 𝑚𝑖,𝑖 and the resonant frequency 𝑓𝑖 can be derived as
𝐹𝐵𝑊∙𝑚𝑖,𝑖 𝐹𝐵𝑊∙𝑚𝑖,𝑖 2
𝑓𝑖 = 𝑓𝑐 ∙ [ + √( ) + 1] (5.15)
2 2
And the normalized self-coupling 𝑚𝑖,𝑖 of resonator i equals to that of the higher or lower band
pass filter.
𝐵𝑊higher / lower band

𝑚𝑖,𝑖 = 1 − (5.16)
𝑓𝑐 ∙𝐹𝐵𝑊
According to the specifications and the coupling matrix synthesis in [9], the normalized
coupling matrix of the diplexer is
{m12 , m23 , m14 , m45 , m11 , m22 , m33 , m44 , m55 , Qe1 , Qe2 , Qe3 } = {0.45, 0.34, 0.45, 0.34, 0, 0.7,
0.74,-0.7,-0.74, 1.34, 2.2, 2.2}.
The normalized coupling matrix is transformed to the desired frequency band with a centre
frequency 𝑓𝑐 = 50 GHz.
The configuration of the proposed diplexer is shown in Fig. 5. 14(a), and different sizes of
resonators are chosen to implement the design. To distribute the input signal to two frequency
bands, the signal from the common resonator to the output resonator must pass through one
output at a particular frequency band while there is little signal passing through to the other
output. For the lower frequency band, the common resonator and the two resonators to Port 2
realize a passband at lower frequencies while the common resonator and the two resonators to
Port 3 realize another passband at higher frequencies. The isolation between Port 2 and 3 should
be high to avoid interference. Since the two coupling paths contribute to two distinct frequency
bands, the isolation is typically good for coupling-based diplexers. The simulation shows that
the diplexer has two frequency bands at 47.5GHz and 52 GHz. This design shows a narrow
passband and only one reflection minimum is realized at each frequency band. To obtain better
filtering performances with more in-band poles, more resonators can be used.
P a g e | 129
(a)
(b)
Fig. 5. 14 (a) Proposed five-resonator diplexer, and (b) its performances.
P a g e | 130
5.5 Fabrication Process and Measurement
Due to the limitation of facilities, the minimum width or gap of metal is 0.1 mm if using LPKF
ProtoMat S62. To prove the design concept, the designs in this Chapter have been scaled down
from 50 GHz to 2 GHz as shown in Fig. 5. 15 for experimental validation. A equal splitting
PD is designed in Sonnet on a Rogers RT6010LM substrate with 𝜀𝑟 = 10.2, and a thickness of
1.27 mm. The smallest width of the gap/strip is 0.2 mm, which is the smallest fabrication
dimension of LPKF ProtoMat S62. Three 2.92mm edge launch connectors are clapped onto
the three ports tightly. The edges of signal lines are wore off finely to prevent short-circuit to
connectors. The circuit is measured by using a VNA (Agilent FieldFox N9917A) with a
frequency range up to 18 GHz. A 50 Ω male load is always attached to the spare port that is
not being measured. The simulated and measured results are imported to Origins for data
plotting as shown in Fig. 5. 16. The measured S11 is better than 20 dB. S21 is about -4 dB, which
means there is 1 dB loss caused by fabrication error and substrate. It can be observed from the
comparison that the measured S-parameters are in good agreement with the simulation.
Unit: mm
(a) (b)
Fig. 5. 15 (a) Circuit layout of a 2 GHz PD, and (b) the fabricated circuit.
P a g e | 131
Fig. 5. 16 Comparison between the simulated and measured results of the 2 GHz PD.
Unit: mm
(a) (b)
Fig. 5. 17 (a) Circuit layout of a 2 GHz diplexer, and (b) the fabricated circuit.
P a g e | 132
(a)
(b)
Fig. 5. 18 Comparison between the simulated and measured (a) S11, S22 and S33, and (b)
S21, S31 and S23 of the 2 GHz Diplexer.
P a g e | 133
A 2 GHz diplexer has been fabricated and measured in the same way. The circuit layout and
the fabricated circuit are shown in Fig. 5. 17. The simulated and measured results are plotted
using Origins as shown in Fig. 5. 18. The measured return loss S11 for lower frequency band is
better than 14 dB. However, the return loss for higher frequency band is less than 7 dB which
is caused by the fabrication precision (±0.1 mm in this case) of the gap between the resonators
for higher frequency band. The fabrication precision also contributes to a relatively high
insertion loss in measurement. In all, the measurement traces agree with simulations. The
measured results show the potential of the proposed method to be applied for mm-Wave PD
and diplexer designs to avoid using T- or Y- junctions and air-bridges.
5.6 Summary
In this chapter, two PDs and a diplexer have been designed by using coupled resonators. The
idea of filter designs has been employed which enables the proposed circuits to have filtering
responses. The bandwidth can be tuned by controlling the coupling coefficients between
resonators. The rectangular G-type spiral resonators have been used to implement all the
designs, allowing significant size miniaturisation. Moreover, a common spiral resonator
replaces the T-junction between two branches to avoid using air-bridges. It should be noted
that with the same distance between resonators, different orientations of two adjacent
resonators will result in different coupling coefficients which in turns produces distinct power
division ratios (1:1, 1:2, 1:4 and 1:8). The PDs occupy very compact sizes. The one with three
resonators has a size of 0.13 𝜆𝑔 × 0.1 𝜆𝑔 while the size of the PD with five resonators is
roughly 0.2 𝜆𝑔 × 0.06 𝜆𝑔 . One equal splitting PD operating at 2 GHz has been fabrication and
measured, the measured S11 is better than 20 dB which has verified the method well.
The diplexer design employs a common resonator to connect two branches to avoid using extra
air-bridges and also eliminate unnecessary loss caused by T- or Y-junctions. The simulation
shows two narrow frequency bands are located at 47.5 GHz and 52 GHz with S11 better than
20 dB. The isolation of these two branches is lower than 20 dB. The circuit size of the diplexer
excluding feed lines is 0.07 𝜆𝑔 × 0.26 𝜆𝑔 which is comparably compact compared to the
designs with connecting junctions. A diplexer with 2 GHz centre frequency has also been
P a g e | 134
fabricated and tested. The measured results show that the diplexer is more sensitive to the gap
between resonators, extra loss is caused by the fabrication precision.
References
[1] Chen, X.P., and Wu, K. ”Substrate integrated waveguide filter: basic design rules and fundamental
structure features,” IEEE Microw. Mag., 2014, 15, (5), pp. 108–116.
[2] Zhang, Q.-L., Yin, W.-Y., He, S., and Wu, L.-S.: ‘Compact substrate integrated waveguide (SIW)
bandpass filter with complementary split-ring resonators (CSRRs)’, IEEE Microw. Wirel. Compon.
Lett, 2011, 20, (8), pp. 426–428.
[3] M. Strackx, K. Janssen, E. D’Agostino, G. Vandenbosch, P. Reynaert, and P. Leroux, “Ultra-
wideband antipodal Vivaldi antenna array with Wilkinson power divider feeding network,” in Proc.
IEEE ICUWB, 2011, pp. 1–4.
[4] H.-M. Hizan, Z. Amback, et al, “Effect of Insertion Losses on Millimeter-Wave SIW Filters Using
LTCC Technology,” 2015 IEEE International RF and Microwave Conference (RFM), Kuching,
Malaysia.
[5] C. P.Wen, “Coplanr waveguide: a surface strip transmission line suitable for nonreciprocal
gyromagnetic device applications”, IEEE Trans. MTT-7, 1969, 1087-1090.
[6] H. Issa, P. Ferrari, E. Hourdakis, and A. G. Nassiopoulou, “On-chip high-performance millimeter-
wave transmission lines on locally grown porous silicon areas,” IEEE Trans. Electron Dev., vol.
58, no. 11, pp. 3720–3724, Nov. 2011.
[7] Tang, X.-L., Franc, A.-L., Pistono, E., Siligaris, A., Vincent, P., Ferrari, P., and Fournier, J,”
Performance improvement versus CPW and loss distribution analysis of slow-wave CPW in 65mm
HR-SOI CMOS technology,” IEEE Trans. Electron Devices, 2012, 59, (5), pp. 1279-1285.
[8] G. Ghione and C.U.Naldi, “Coplanar waveguide for MMIC applications: Effect of upper shielding,
conductor backing, finite extent ground planes, and line-to-line coupling”, IEEE Trans. MTT-35,
1987, 260-267.
[9] Hong, J.S.G. and Lancaster, M.J., 2004. Microstrip filters for RF/microwave applications (Vol.
167). John Wiley & Sons.
[10] P. Vélez, J. Selga, J. Bonache, and F. Martín, “Slow-wave inductivelyloaded electromagnetic
bandgap (EBG) coplanar waveguide (CPW) transmission lines and application to compact power
dividers,” in Proc. Eur. Microw. Conf. (EuMC), Oct. 2016, pp. 104–107.
[11] Tiku Yu, "A Broadband Wilkinson Power Divider Based on the Segmented Structure", Microwave
Theory and Techniques IEEE Transactions on, vol. 66, no. 4, pp. 1902-1911, 2018.
P a g e | 135
[12] C. Li, , L. B. Lok, A. Khalid, V. Papageorgiou, J. Grant, and D. R. S. Cumming, “A coplanar ring
power divider with high isolation for Vband and W-band applications,” Proceedings of the 42nd
European Microwave Conference, 57–60, 2012.
[14] E.-M. Fall, F.-Domingue, and A. Reinhardt, “Design of RF Power Combiner/Divider based on
Surface Acoustic Wave Longitudinally Coupled Resonators,” 2017 IEEE International
Ultrasonics Symposium (IUS), Washington, DC, USA.
[15] J. Zhou, H, Qian, D. Huang and X. Luo, “Ultra-Wideband (UWB) Wilkinson Power Divider With
Ultra-Narrow Dual-Notched Bands Using Embedded CPW Resonators,” 2017 IEEE MTT-S
International Microwave Symposium (IMS), Honololu, HI, USA.
[16] G. Leon, L.-F. Herran and F.-L. Heras, “Novel CPW Power Divider Based on Resonators,”
European Conference on Antennas and Propagation (EUCAP), 2012 6th. Prague, Czech Republic.
[17] Ma, Z, Kawaguchi, T, Kobayashi, Y, Koizumi, D, Satoh, K, Narahashi, S.”A miniaturized high
temperature superconducting bandpass filter using CPW quarter-wavelength spiral resonators’.
IEEE MTT-S, Int. Microwave Symp. Digest, June 2006, pp. 1197–1200.
[18] Z. Xu, “A compact dual-band bandpass superconducting filter using microstrip/CPW spiral
resonators,” IEEE Microw. Wireless Compon. Lett. vol. 23, no. 11, pp. 584–586, Nov. 2014.
[19] K. Song, X. Ren, F. Chen, and Y. Fan, “Compact in-phase power divider integrated filtering
response using spiral resonator,” IET Microw. Antennas Propag., vol. 8, no. 4, pp. 228–234, 2014.
[20] T. Zheng et al., “Design of compact superconducting diplexer with spiral short-circuited stubs,”
IEEE Trans. Appl. Supercond., vol. 24, no. 2, Apr. 2014, Art. ID. 1500405.
[21] T. N. Zheng et al., “Compact superconducting diplexer design with conductor-backed coplanar
waveguide structures,” IEEE Trans. Appl. Supercond., vol. 25, no. 2, Apr. 2015, Art. no. 1501304.
[22] S. Bastioli, L. Marcaccioli, and R. Sorrentino, “An original resonant Y-junction for compact
waveguide diplexers,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 7–12, 2009, pp. 1233–1236.
[23] C. E. Saavedra, “Diplexer Using a Circulator and Interchangeable Filters,” Proceeding of the 7th
International Caribbean Conference on Devices, Circuits and Systems, Mexico, 28-30 April 2008,
pp. 1-5. doi:10.1109/ICCDCS.2008.4542631.
[24] X. Guan et al., “Compact, low insertion-loss, and wide stopband HTS diplexer using novel
coupling diagram and dissimilar spiral resonators,” IEEE Trans. Microw. Theory Techn., vol. 64,
no. 8, pp. 2581–2589, Aug. 2016.
[25] A. O. Nwajana, and K. S. K. Yeo, “Microwave diplexer purely based on direct synchronous and
asynchronous coupling,” Radio engineering, vol. 25, no. 2, pp. 247-252, Jun. 2016.
P a g e | 136
[26] Xia, W. (2015). Diplexers and Multiplexers Design by Using Coupling Matrix Optimisation. Ph.D
Thesis, University of Birmingham. Etheses.bham.ac.uk.
P a g e | 137
Chapter 6: Summary and Future Work
Chapter 6 Summary and Future Work
6.1 Summary
To summarise the contents of this thesis, it has been demonstrated that adding capacitors in
parallel with transmission lines can effectively adjust the operating frequency of PDs. By
carefully choosing the PD topology and the impedance of the transmission lines, broadband
and reconfigurable responses with a wide tuning range can be achieved. Experiments have
validated the design methods.
The other research objective is mm-wave PDs. Due to some common challenges in mm-Wave
device design, such as high loss and low tolerance of fabrication error, the development of mm-
Wave PDs is very challenging. Based on the improvement of manufacturing techniques, many
simple but effective designs are promising.
Several key challenges regarding PD designs are listed below:
1) Due to the increasing number of multi-standard systems, PDs with multi-band, broadband
or reconfigurable features have been investigated in depth. Although many published papers
provide plenty of comprehensive solutions, most of them are hybrid technologies that require
several essential design elements, such as using both coupled-line and stepped-impedance
transformers to achieve wideband performance. Hybrid solutions are good choices for specific
applications because they accomplish specifications. However, more versatile but
straightforward designs are the most desired, because a simple design is more desired to be
integrated into systems with minimised side effects.
2) As the competition among communication industries increases the scarcity of radio spectrum
resource, PDs at higher frequencies, from mm-Wave to even Terahertz bands, are in high
demand. Feasible solutions need to be investigated in depth.
3) Most high frequency-related researches are focusing on waveguide structures as they

perform low loss and high-power capability features. However, they usually suffer from the
large size and relatively high cost. Moreover, conventional waveguides are not perfect choices
P a g e | 138
for circuit integration. As a result, planar designs working on such high frequencies are
imperatively desired. Since the majority of planar PDs are designed based on transmission lines,
they usually have relatively high loss for high-frequency applications. PDs with a simple layout
and a low-cost precision fabrication process are of great value to be investigated.
6.2 Key Contributions
This thesis has provided a thorough study of microwave PD designs and also presented a
promising mm-Wave PD design.
The key contributions are detailed as follows.
Chapter 2
1) A comprehensive literature review has been carried out to show the critical and recent
development of broadband and reconfigurable PD designs in microwave communication
systems. It has also depicted several available mm-Wave PD designs to demonstrate the
design and fabrication challenges.
Chapter 3
2) The idea of adding capacitors in parallel with main transmission lines is firstly proposed.
This proposed structure solves three main challenges. Firstly, the added capacitors have
broadened the overall bandwidth to a great extent (2.7:1 of 15-dB bandwidth for a two-
way prototype). Secondly, changing the capacitance can result in a tunable bandwidth. The
last key point is from the physical aspect of view. The proposed structure has two short
stubs connecting the output ports which adding the physical isolation between them.
3) Also, an eight-way PD is designed and fabricated by cascading two-way structures
together. The measured results indicate that the idea can be adapted to multi-way
broadband PD designs which validates the importance of this work.
Chapter 4
4) By adding a pair of the capacitor in parallel with the main transmission lines, and by
carefully choosing the lengths and impedances of the transmission lines, the operational
frequency of the PD can be tuned over a wide range. The design in this Chapter focused
on the main transmission lines and the stubs that connect the main transmission lines and
the lumped-elements RLC in the isolation circuit. An analytic solution is provided in the
P a g e | 139
chapter to predict and control the frequency tuning range and its return loss performance.
The designed two-way circuit uses three varactors and their corresponding bias circuits to
tune the frequency band electrically. Measurement shows that the centre frequency with
better than 20 dB return loss can be tuned from 0.9 GHz to 4.2 GHz, which is the widest
compared to other similar works.
Chapter 5
5) The work in chapter 5 focuses on developing an mm-Wave PD using rectangular spiral
resonators. Using a spiral resonator will significantly reduce the occupied area of the
circuit. Besides, using a common resonator near the input port can achieve a low loss
performance and also size reduction compared the traditional design using a T- or Y-
junction and air-bridges. A CPW structure has its signal line and ground on the same side
of the substrate which reduces the complexity of fabrication and makes it easier for
integration. The design can effectively control the power division by changing the
distances between the resonators. Moreover, a diplexer has been implemented by
appropriately choosing the size and coupling coefficients of the resonators. Good
demonstration has been made by testing frequency scaled-down PD and diplexer.
6.3 Future Work
Based upon the summary above, and considering the challenges of existing technologies,
further works can be carried out in the following areas.
1) In Chapter 5, a 2 GHz PD and a diplexer have already been designed and tested to validate
the design theory due to facility limitation. PDs and diplexer operating at 50 GHz can be
fabricated and measured by using following method.
The PD and diplexer will employ the fabrication process in [1] on a 625 µm thick semi-
insulating GaAs substrate with a loss tangent of 0.006. All patterns are defined using Electron
beam lithography (EBL) which is a powerful technique for creating nanostructures that are too
small to fabricate with conventional photolithography.
P a g e | 140
Fig. 6. 1 Resist pattern schematic by using EBL [2].
The technique works by moving a highly focussed electron beam [2] over a sample to write
out a pattern designed with Computer-aided Design (CAD) tools. The pattern is recorded in an
electron sensitive film (or resist) deposited on the sample before exposure by spin coating. The
electron beam induces a change in the molecular structure and solubility of the resist film.
Following exposure to the electron beam as shown in Fig. 6. 1, the resist is developed in a
suitable solvent to selectively dissolve either the exposed or unexposed areas of the resist.
There are two main reasons for using EBL. One is precision. EBL process can create structures
at the level of 10 nm.
The device will be tested using on-wafer probes and a VNA system as shown in Fig. 6. 2(c).
A typical 110 GHz system contains a VNA, cables, probes, probe positioners, probe station
contact substrate, calibration substrate, calibration software, bias supply and a microscope [3].
The components will be tested using a pair of DC-110 GHz Ground-Signal-Ground (GSG)
probes (as shown in Fig. 6. 2(a)(b)) with a 100 μm pitch separation from GGB Industries and
an Agilent PNA N5250C on a Cascade semi-automated probe station. CPW transitions from
20 µm /15 µm to 60 µm /40 µm will be deployed at three ports to be compatible with the
geometry of the test probes. A third probe terminated by a broadband 50 Ω load on the 1 mm
coaxial port side is used to load the idle port [4]. This three-port component test method, unlike
the conventional method that requires three devices of the same design but terminated with 50
Ω thin-film resistors for different ports [5], allows the same Device Under Test (DUT) to be
characterised using a two-port VNA. Off-wafer calibration using a calibration substrate CS-15
from GGB Industries will used to calibrate the system reference plane to the probe tips before
taking measurements.
P a g e | 141
(a) (b)
(c)
Fig. 6. 2 (a) The GSG probe, and (b) microphotograph of the fabricated ring divider
under test [6], (c) Typical 110 GHz System [7].
The calibration will employ the Short-Open-Load-Reciprocal Thru (SOLR) method [8]-[10]
for Port 1 to Port 2/Port 3, and a Line-Reflect-Reflect-Match (LRRM) calibration method will
be used for Port 2 to Port 3 isolation measurement for it is not sensitive to the load asymmetry
[11]. SOLR uses the SOLT Short- Open-Load Thru (SOLT) short, open, and load standards
and requires the same characterization knowledge. SOLR calibration does require that the
P a g e | 142
through standard be reciprocal, such that S21 equals S12. The probe station measurement
requires a circuit configuration of input port to be orthogonal to the directions of output ports.
Most circuits would not strictly satisfy the measurement requirements so that calibration is
essential in order to obtain accurate results. The limited knowledge requirement on the through
standard makes SOLR calibration ideally suited for situations where a non-ideal through
standard is desired, such as with on-wafer probing where the probes are positioned 90° from
each other (orthogonally) during the desired measurement. A through standard with a 90° bend
could be used for calibration. A bend makes the through behave non-ideal and therefore creates
errors or degradation in those calibration models relying on the ideal characteristics. SOLR
calibration can overcome this limitation since it does not rely on an ideal through model so that
eliminate degradation during the measurement. The use of the SOLR algorithm with a 90°
through standard simplifies both calibration and measurement, which provides the best
calibration for these less ideal measurement configurations.
Based on the results in Chapter 3, the topology is promising to be used in mm-wave PD

designs to extend the operational bandwidth. However, surface mount capacitors are not
widely available at such high frequencies. An interdigital capacitor as shown in
2) Fig. 6. 3 may be a good candidate as a semi-lumped element for mm-Wave device designs
[12]-[13]. The design objectives are to provide the desired capacitance at the design
frequency in a reasonably small area. The interdigital capacitors are easy to fabricate and
have more freedom to achieve desired capacitance due to their structures.
Fig. 6. 3 Configuration of an interdigital capacitor.

P a g e | 143
3) Since reconfigurable devices become more and more popular, it is of great value to
investigate the conept in more depth. The term “reconfigurable” here may refer to different
types of response instead of the centre frequency, bandwidth or power division ratio. For
example, in Chapter 5, the PD that was designed based on coupling structures can be
switched to a diplexer. Many papers on reconfigurable operating frequency [14]-[17] have
been published. They are based on integrating varactors to the circuit to continuously
control the resonant frequencies changing in a certain range. As the work in [6] indicates,
the external quality factor Qe and the operating frequency can be adjusted by altering the
varactors on strait- and ring-slots. As a result, the two frequency bands can be controlled
synchronously. Thus, a PD and diplexer can be switchable if the circuit configurations and
bias circuits are carefully designed.
References
[1] C. Li, A. Khalid, V. Papageorgiou, L. B. Lok, and D. R. S. Cumming, “A wideband CPW ring
power combiner with low insertion loss and high port isolation,” IEEE Microw. Wireless Compon.
Lett., vol. 22, no. 11, pp. 580–582, Nov. 2012.
[2] Sheffield, U. (2018). Overview - EBL Patterning - Electron Beam Lithography - The University
of Sheffield. [online] Sheffield.ac.uk. Available at: https://www.sheffield.ac.uk/ebl/patterning
[Accessed 16 Oct. 2018].
[3] Fisher, G. (2018). A guide to Successful on Wafer Millimeter wave Rf characterisation. [online]
Keysight.com. Available at:
https://www.keysight.com/upload/cmc_upload/All/OnWaferMillimeter.pdf [Accessed 16 Oct.
2018].
[4] C. Li, et al., “Novel on-wafer measurement technique for passive multiport devices in millimetre
wave frequency range”, in 9th Millimetre-wave User Group Meeting. Unpublished, p. 2, Glasgow,
UK. March 2011.
[5] D. Lopez-Diaz, et al., “Monolithic integrated 210 GHz couplers for balanced mixers and image
rejection mixers”. Workshop on Integrated Nonlinear Microwave and Millimeter-Wave Circuits
(INMMIC), 2010: p.4.
P a g e | 144
[7] Fisher, G. (2018). A guide to Successful on Wafer Millimeter wave Rf characterisation. [online]
Keysight.com. Available at:
https://www.keysight.com/upload/cmc_upload/All/OnWaferMillimeter.pdf [Accessed 16 Oct.
2018].
[8] S. Basu and L. Hayden, “An SOLR calibration for accurate measurement of orthogonal on-wafer
DUTs,” IEEE MTT-S Digest, Denver CO, Jun 1997, pp. 1335-1338.
[9] Keysight Agilent Technologies. (2006). Advanced Calibration Techniques for Vector Network
Analyzers. [online] Available at: http://anlage.umd.edu/Agilent_Advanced_VNA_calibration.pdf
[10] Zone.ni.com. (2018). Short-Open-Load-Reciprocal (SOLR) Calibration - NI Vector Network
Analyzers (NI-VNA 14.0) Help - National Instruments. [online] Available at:
http://zone.ni.com/reference/en-XX/help/373153D-01/vnahelp/calibration_solr/ [Accessed 15 Oct.
2018].
[11] Citeseerx.ist.psu.edu. (2018). A Review of VNA Calibration Methods. [online] Available at:
http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.662.2977&rep=rep1&type=pdf
[12] F. Aryanfar and K. Sarabandi, “Characterization of semilumped CPW elements for millimetre-
wave filter design,” IEEE Tran. Microwave Theory Techn., vol. 53, no. 4, pp. 1288–1293, Apr.
2005.
[13] F. Aryanfar and K. Sarabandi, “Compact millimeter-wave filters using distributed capacitively
loaded CPW resonators,” IEEE Trans. Microw. Theory Tech., vol. 54, no. 3, pp. 1161–1165, Mar.
2006.
[14] G. Zhang, Z. Qian and J. Yang, “Design of a compact microstrip power-divider diplexer with
simple layout,” Electronic Letters, vol: 4, issue:16, pp.1007-1009, Aug. 2018.
[15] M. S. Sorkherizi, A. Vosoogh, A. A. Kishk, and P.-S. Kildal, “Design of integrated diplexer-power
divider,” IEEE MTT-S Int. Microw. Symp. Dig., May 2016, pp. 1–3.
[16] P. Deng, W. Lo, B-L. Chen and C.-H. Lin, “Designs of Diplexing Power Dividers,” IEEE Access,
vol:6, pp. 3872-3881, Jan. 2018.
[17] X. Liu, Z. Tian, H-Jenny Qian, et al, “Reconfigurable Diplexer Using λ/2 Resonator with Hybrid
Varactor Embedded Stepped-Impedance Open-Stub and Slot,” IEEE MTT-S International
Wireless Symposium (IWS), Chengdu China, May 2018.
P a g e | 145
END
P a g e | 146

WPD Dissertation

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

WPD Dissertation

Uploaded by

Copyright:

Available Formats

Investigation of Planar Power Dividers

for Wireless Communication Systems

A thesis submitted in accordance with the requirements for the award of

the degree of Doctor of Philosophy of the University of Liverpool

Table of Contents .................................................................................................................... iv

Fig. 1. 1. A diagram of wireless communication networks [1]. ......................................................................... 1

Table 3. 1 COMPARISON OF CALCULATED AND OPTIMIZED PARAMETERS. ............................. 72

[1] A. Chen, Y. Zhuang, J. Zhou, Y. Huang and L. Xing, “Design of a Broadband

CAD Computer-aided Design

CDMA Code Division Multiple Access

CPS Symmetric Coplanar Strip-lines

CPW Coplanar Waveguide

CRLH Composite Right/Left Handed

DUT Device Under Test

EBL Electron Beam Lithography

GSM Global System for Mobile Communications

HSPA High-Speed Packet Access

LTCC Lower Temperature Co-Fired Ceramic

LTE Long-Term Evolution

MHMIC Monolithic Hybrid Microwave Integrated Circuit

MMIC Monolithic Microwave Integrated Circuit

OFDM Orthogonal Frequency-Division Multiplexing

PCB Printed Circuit Boards

RFID Radio Frequency Identification

SIW Substrate Integrated Waveguide

TDMA Time-Division Multiple Access

VNA Vector Network Analyzer

Fig. 1. 1. A diagram of wireless communication networks [1].

1.1 Power Dividers

Fig. 1. 4. Transmission line model of a lossless T-junction.

⇒ A lossless divider has mismatched ports

Fig. 1. 5. N-way Wilkinson PD [6].

A two-way Wilkinson PD is relatively simple and can be realised by different technologies

Furthermore, contributed by the isolation circuit realised by resistors, high-level isolation is

Fig. 1. 8. Basic Gysel topology.

1.2 PD in Modern Wireless Communication Systems

In general, the main preferred functionalities of a PD can be summarized as the following

1.3 The Motivations of the Work

 To study the wideband PD technologies and investigate the mechanisms of these

 To design an mm-Wave coplanar waveguide PD using coupled resonators for easy

1.4 Thesis Organization

The content of this thesis are organized in the following manner.

[1] Hilink.com.cn. (2018). Reactive Power Splitter-Power Splitter-Manufacturer of Passive Products.

Chapter 2 A Review of Power Dividers

2.1 Broadband PD Technologies

2. 1. 1 PDs with Multi-section

Theoretically, the conventional single-section two-way Wilkinson-type PD has a usable

2. 1. 2 PDs with Stepped-impedance Transformers

Fig. 2. 4 Structure of stepped-impedance transformer.

Papers [44]-[45] propose an impedance-matching method of interconnecting transmission lines

2. 1. 3 PDs with Coupled-Line

Fig. 2. 6 (a) Schematic model of coupled-line, (b) coupled-line side view.

Conventional two-way Wilkinson PDs consists of two quarter-wavelength lines and an

2.1.3 Hybrid and Multi-Functional Wideband PDs

Fig. 2. 9 (a) Schematic of proposed Ultra-wideband PD and (b) fabricated circuit.

A multi-functional, ultra-wideband PD using coupled-line is reported in [69]. This design

2.2 Reconfigurable PDs

2. 2. 1 Tunable Power Dividing Ratios

2. 2. 2 Tunable Operating Frequency

Fig. 2. 13 (a) Configuration and (b) fabricated circuit of the PD.

2. 2. 3 Hybrid Reconfigurable PDs

2.3 Coupled Resonator PDs

A compact and high in-band isolation resonator-based PD with a second-order Chebyshev

2.4 Millimetre-Wave PDs